CN105608448B - A method and device for extracting LBP features based on facial key points - Google Patents

Abstract

The invention discloses a method and device for extracting LBP features based on facial key points. The invention relates to the technical field of image processing and pattern recognition, and aims to solve the problem of extracting the same key point of the face under the same person and different face poses. The problem of large differences in LBP features can improve the accuracy of face recognition. This method determines the face pose of the face image according to the key points of the face image, and then corrects the ellipse neighborhood radius corresponding to different face key points according to the face pose of the face image, and establishes the relationship between the ellipse neighborhood radius and The corresponding relationship between the face poses of the face image avoids the problem that the local zoom ratio of the face image is different under different face poses of the same person, resulting in a large difference in the LBP features extracted from the same key point, and improves the accuracy of different face images. The effectiveness of LBP feature extraction based on facial key points under face pose, thereby improving the accuracy of face recognition.

Description

A method and device for extracting LBP features based on facial key points

technical field

The invention relates to the technical fields of image processing and pattern recognition, in particular to a method and device for extracting LBP features based on facial key points.

Background technique

Face recognition technology is to identify the identity by analyzing the displacement shape and positional relationship of facial organs. It is an important biometric technology and is widely used in security, access control, monitoring and other fields. The main algorithms of face recognition technology include face recognition methods based on template matching based on geometric features, face recognition methods based on geometric features, face recognition methods based on sample learning, and face recognition methods based on texture features. Among them, the face recognition method based on facial texture features mainly relies on LBP (Local Binary Pattern), that is, local binary pattern for facial feature extraction.

The face feature extraction method based on LBP mainly includes two types, one is to divide the entire face image into blocks, perform multi-scale LBP feature extraction on each block image, and combine the LBP features of all blocks after normalization to obtain the whole face image. The LBP feature of the face; the other is based on the key points of the face (for example, eyeballs, nose, corners of the mouth, eyebrows, etc.), extracting the LBP features of a certain area around it, and normalizing the LBP of all key points The features are combined to obtain the LBP features of the entire face. However, the above two facial feature extraction methods based on LBP, for the facial images of the same person with different postures, all use a certain point as the center point to collect the pixel values of several sampling points in a circular neighborhood with a fixed radius , and then compare the pixel values of the several sampling points with the pixel values of the center point to obtain the LBP feature value of the center point. Therefore, the existing facial feature extraction method based on LBP does not consider the difference in facial image scaling caused by posture changes, resulting in the facial images of the same person under different postures, for the central point of the same position, the extracted The pixel values of several sampling points in the circular neighborhood of the same radius are different, resulting in a large difference in the LBP features of the center point, which in turn leads to inaccurate face recognition.

Exemplary, as shown in accompanying drawing 1 and accompanying drawing 2, shown in Fig. 1 and Fig. 2 is the face image of the same person's front face and left side face respectively, for the LBP feature extraction of the same central point A at the corner of its mouth, First, extract the pixel values of N sampling points in the circular neighborhood with radius R shown in Figure 1 and Figure 2. Referring to Figure 1 and Figure 2, due to the left face, N sampling points in the circular neighborhood with radius R Some of the sampling points exceed the range of the face, resulting in a large difference in the pixel values of the N sampling points in the same circular neighborhood of the same center point A shown in Figure 1 and Figure 2, which in turn leads to the The LBP features of the same center point A at the proposed location are quite different, which leads to the fact that the facial images shown in Figure 1 and Figure 2 cannot be recognized as the same person, resulting in inaccurate face recognition.

Contents of the invention

The embodiment of the present invention provides a face feature extraction method and device based on LBP, which aims to solve the problem that the extracted LBP features of the same key point of the face are quite different under different face postures of the same person, thereby improving face recognition. Accuracy.

The concrete technical scheme that the present invention provides is as follows:

A method for extracting LBP features based on facial key points, comprising:

Estimating the facial pose of the facial image based on key points of the facial image to be processed;

Correcting the radius of the ellipse neighborhood corresponding to the key point according to the face pose of the face image;

Using the ellipse neighborhood radius to perform LBP feature extraction on the key points on the face image.

Further, the estimation of the facial pose of the facial image based on the key points of the facial image to be processed includes:

Obtaining a first vertical distance and a second vertical distance in the vertical direction of the facial image, wherein the first vertical distance and the second vertical distance are distributed from top to bottom in the vertical direction;

If the ratio between the first vertical distance and the second vertical distance is greater than a first threshold, then the facial posture of the facial image is head down; or the first vertical distance and the second vertical distance If the ratio between the vertical distances is smaller than the first threshold, the facial posture of the facial image is head up.

Further, the estimation of the facial pose of the facial image based on the key points of the facial image to be processed further includes:

Obtaining a first horizontal distance and a second horizontal distance in the horizontal direction of the facial image, wherein the first horizontal distance and the second horizontal distance are distributed from left to right along the horizontal direction;

If the ratio between the first horizontal distance and the second horizontal distance is greater than a second threshold, then the face posture of the facial image is the right head; or the first horizontal distance and the second horizontal distance If the ratio between the distances is smaller than the second threshold, the face pose of the face image is the left head.

Preferably, the modifying the radius of the ellipse neighborhood corresponding to the key point according to the facial pose of the facial image includes:

If the face posture is a front face, the horizontal axis of the ellipse neighborhood is equal to the first preset value, and the vertical axis of the ellipse neighborhood is equal to the first preset value; or if the face posture is head-up Or lower your head, the horizontal axis of the ellipse neighborhood is equal to the first preset value, and the vertical axis of the ellipse neighborhood is equal to the product of the first preset value and the first preset weight; or if the facial posture is left For a side head or a right head, the horizontal axis of the ellipse neighborhood is equal to the product of the first preset value and the second preset weight, and the vertical axis of the ellipse neighborhood is equal to the first preset value.

Preferably, the face image includes a first area, a second area, a third area and a fourth area, wherein the first area, the second area, the third area and the fourth area Distributed clockwise, and the first area is located in the vertical direction of the fourth area.

Further, the modifying the radius of the ellipse neighborhood corresponding to the key point according to the face pose of the face image includes:

If the face posture is a frontal face, the horizontal axis of the ellipse neighborhood and the vertical axis of the ellipse neighborhood in the first area, the second area, the third area, and the fourth area The axes are all equal to the first preset value;

If the facial posture is head-up, the horizontal axis of the first area, the second area, the third area and the fourth area is equal to the first preset value, and the first The vertical axes of the first area and the second area are smaller than a first preset value, and the vertical axes of the third area and the fourth area are larger than the first preset value;

If the facial posture is head down, the horizontal axes of the first area, the second area, the third area and the fourth area are equal to the first preset value, and the first The vertical axes of the zone and the second zone are greater than a first preset value, and the vertical axes of the third zone and the fourth zone are smaller than the first preset value;

If the face posture is the left head, the horizontal axis of the first area and the third area is greater than the first preset value, and the vertical axis of the second area and the fourth area less than the first preset value, the vertical axes of the first area, the second area, the third area and the fourth area are equal to the first preset value;

If the face posture is the right head, the horizontal axis of the first area and the third area is smaller than the first preset value, and the vertical axis of the second area and the fourth area greater than the first preset value, the vertical axes of the first area, the second area, the third area and the fourth area are equal to the first preset value;

If the face posture is head up and left head, the horizontal axis of the first area and the third area is greater than the first preset value, and the second area and the fourth area The horizontal axis of is smaller than the first preset value, the vertical axes of the first area and the second area are smaller than the first preset value, and all of the third area and the fourth area The vertical axis is greater than the first preset value;

If the face posture is head up and head on the right side, the horizontal axis of the first area and the third area is smaller than the first preset value, and the second area and the fourth area The horizontal axis of is greater than the first preset value, the vertical axis of the first area and the second area is greater than the first preset value, and all of the third area and the fourth area The vertical axis is smaller than the first preset value;

If the face posture is head down and left head, the horizontal axis of the first area and the third area is greater than the first preset value, and the horizontal axis of the second area and the fourth area The horizontal axis is smaller than the first preset value, the vertical axes of the first area and the second area are larger than the first preset value, the third area and the fourth area are the vertical axis is smaller than the first preset value;

If the face posture is head up and left head, the horizontal axis of the first area and the third area is smaller than the first preset value, and the second area and the fourth area The horizontal axis of is greater than the first preset value, the vertical axes of the first area and the second area are smaller than the first preset value, and all of the third area and the fourth area The vertical axis is greater than the first preset value.

On the other hand, the present invention also provides a method for extracting LBP features based on facial key points, including:

Estimating the facial pose of the facial image based on key points of the facial image to be processed;

determining the radius of the ellipse neighborhood corresponding to the key point in the preset mapping table according to the face pose of the face image;

Using the ellipse neighborhood radius to perform LBP feature extraction on the key points on the face image.

Further, the estimation of the facial pose of the facial image based on the key points of the facial image to be processed includes:

Obtaining a first vertical distance and a second vertical distance in the vertical direction of the facial image, wherein the first vertical distance and the second vertical distance are distributed from top to bottom in the vertical direction;

If the ratio between the first vertical distance and the second vertical distance is greater than a first threshold, then the facial posture of the facial image is head down; or the first vertical distance and the second vertical distance If the ratio between the vertical distances is smaller than the first threshold, the facial posture of the facial image is head up.

Further, the estimation of the facial pose of the facial image based on the key points of the facial image to be processed further includes:

Obtaining a first horizontal distance and a second horizontal distance in the horizontal direction of the facial image, wherein the first horizontal distance and the second horizontal distance are distributed from left to right along the horizontal direction;

If the ratio between the first horizontal distance and the second horizontal distance is greater than a second threshold, then the face posture of the facial image is the right head; or the first horizontal distance and the second horizontal distance If the ratio between the distances is smaller than the second threshold, the face pose of the face image is the left head.

Preferably, the face image includes a first area, a second area, a third area and a fourth area, wherein the first area, the second area, the third area and the fourth area Distributed clockwise, and the first area is located in the vertical direction of the fourth area.

Preferably, the preset mapping table is used to characterize the ellipse neighborhood radius corresponding to the key point in different face regions under different face poses, wherein:

If the face posture is a frontal face, the horizontal axis of the ellipse neighborhood and the vertical axis of the ellipse neighborhood in the first area, the second area, the third area, and the fourth area The axes are all equal to the first preset value;

If the facial posture is head-up, the horizontal axis of the first area, the second area, the third area and the fourth area is equal to the first preset value, and the first The vertical axes of the first area and the second area are smaller than a first preset value, and the vertical axes of the third area and the fourth area are larger than the first preset value;

If the facial posture is head down, the horizontal axes of the first area, the second area, the third area and the fourth area are equal to the first preset value, and the first The vertical axes of the zone and the second zone are greater than a first preset value, and the vertical axes of the third zone and the fourth zone are smaller than the first preset value;

If the face posture is the left head, the horizontal axis of the first area and the third area is greater than the first preset value, and the vertical axis of the second area and the fourth area less than the first preset value, the vertical axes of the first area, the second area, the third area and the fourth area are equal to the first preset value;

If the face posture is the right head, the horizontal axis of the first area and the third area is smaller than the first preset value, and the vertical axis of the second area and the fourth area greater than the first preset value, the vertical axes of the first area, the second area, the third area and the fourth area are equal to the first preset value;

If the face posture is head up and left head, the horizontal axis of the first area and the third area is greater than the first preset value, and the second area and the fourth area The horizontal axis of is smaller than the first preset value, the vertical axes of the first area and the second area are smaller than the first preset value, and all of the third area and the fourth area The vertical axis is greater than the first preset value;

If the face posture is head up and head on the right side, the horizontal axis of the first area and the third area is smaller than the first preset value, and the second area and the fourth area The horizontal axis of is greater than the first preset value, the vertical axis of the first area and the second area is greater than the first preset value, and all of the third area and the fourth area The vertical axis is smaller than the first preset value;

If the face posture is head down and left head, the horizontal axis of the first area and the third area is greater than the first preset value, and the horizontal axis of the second area and the fourth area The horizontal axis is smaller than the first preset value, the vertical axes of the first area and the second area are larger than the first preset value, the third area and the fourth area are the vertical axis is smaller than the first preset value;

If the face posture is head up and left head, the horizontal axis of the first area and the third area is smaller than the first preset value, and the second area and the fourth area The horizontal axis of is greater than the first preset value, the vertical axes of the first area and the second area are smaller than the first preset value, and all of the third area and the fourth area The vertical axis is greater than the first preset value.

In one aspect, the present invention also provides a kind of LBP feature extraction device based on facial key points, comprising:

A key point detection module for detecting key points of the face image to be processed;

An image processing module, configured to obtain the distance between the key points, determine the face pose of the face image according to the distance, and modify the radius of the ellipse neighborhood corresponding to the key point according to the face pose;

A feature extraction module, configured to perform LBP feature extraction on the key points using the radius of the ellipse neighborhood on the face image.

Further, the image processing module is specifically used for:

Acquiring a first vertical distance and a second vertical distance in the vertical direction of the facial image, wherein the first vertical distance and the second vertical distance are distributed from top to bottom in the vertical direction, If the ratio between the first vertical distance and the second vertical distance is greater than a first threshold, then the facial posture of the facial image is head down, or the first vertical distance and the second vertical distance The ratio between the vertical distances is less than the first threshold, then the face posture of the facial image is head-up;

Acquiring a first horizontal distance and a second horizontal distance in the horizontal direction of the facial image, wherein the first horizontal distance and the second horizontal distance are distributed from left to right in the horizontal direction, and the first horizontal distance If the ratio between the first horizontal distance and the second horizontal distance is greater than the second threshold, then the facial posture of the face image is the right head, or the ratio between the first horizontal distance and the second horizontal distance is less than The second threshold, then the face pose of the face image is the left head.

Further, the image processing module is specifically used for:

If the face pose is a frontal face, determine the horizontal axis of the ellipse neighborhood and the ellipse neighborhood of the first area, the second area, the third area, and the fourth area both vertical axes are equal to the first preset value; or

If the facial posture is head-up, it is determined that the horizontal axis of the first area, the second area, the third area, and the fourth area is equal to the first preset value, and the The vertical axes of the first area and the second area are smaller than a first preset value, and the vertical axes of the third area and the fourth area are larger than the first preset value; or

If the facial posture is head down, determine that the horizontal axis of the first area, the second area, the third area, and the fourth area is equal to the first preset value, and the second area The vertical axes of the first area and the second area are greater than a first preset value, and the vertical axes of the third area and the fourth area are smaller than the first preset value; or

If the face posture is the left head, it is determined that the horizontal axis of the first area and the third area is greater than the first preset value, and the vertical axis of the second area and the fourth area axis is smaller than said first preset value, said longitudinal axis of said first zone, said second zone, said third zone and said fourth zone is equal to said first preset value; or

If the face posture is the head on the right side, it is determined that the horizontal axis of the first area and the third area is smaller than the first preset value, and the vertical axis of the second area and the fourth area the axes are greater than the first preset value, and the longitudinal axes of the first zone, the second zone, the third zone, and the fourth zone are equal to the first preset value; or

If the face posture is head up and left head, it is determined that the horizontal axis of the first area and the third area is greater than the first preset value, and the second area and the fourth area The horizontal axis of the region is smaller than the first preset value, the vertical axes of the first region and the second region are smaller than the first preset value, and the third region and the fourth region the vertical axis is greater than a first predetermined value; or

If the face posture is head up and head on the right side, it is determined that the horizontal axis of the first area and the third area is smaller than the first preset value, and the second area and the fourth area The horizontal axis of the area is greater than the first preset value, the vertical axis of the first area and the second area is greater than the first preset value, and the third area and the fourth area the vertical axis is less than a first predetermined value; or

If the face posture is head down and left head, it is determined that the horizontal axis of the first area and the third area is greater than the first preset value, and the second area and the fourth area The horizontal axis of is smaller than the first preset value, the vertical axes of the first area and the second area are greater than the first preset value, and all of the third area and the fourth area said vertical axis is less than a first predetermined value; or

If the face posture is head up and left head, it is determined that the horizontal axis of the first area and the third area is smaller than the first preset value, and the second area and the fourth area The horizontal axis of the region is greater than the first preset value, the vertical axes of the first region and the second region are smaller than the first preset value, and the third region and the fourth region The vertical axis is greater than a first preset value.

The beneficial effects of the present invention are as follows:

The face key point-based LBP feature extraction method provided by the embodiment of the present invention first determines the face pose of the face image according to the key points of the face image, and then corrects different face key points according to the face pose of the face image. The radius of the ellipse neighborhood corresponding to the point establishes the correspondence between the radius of the ellipse neighborhood and the face pose of the face image, thereby avoiding the different face image local scaling ratios of the same person under different face poses, resulting in The problem of large differences in LBP features extracted from the same key point improves the effectiveness of LBP feature extraction based on facial key points under different face poses, thereby improving the accuracy of face recognition.

Description of drawings

Fig. 1 is a schematic diagram of the LBP feature extraction neighborhood of the facial key point A under the front face posture in the prior art;

Fig. 2 is a schematic diagram of the LBP feature extraction neighborhood of the face key point A under the profile pose in the prior art;

Fig. 3 is a kind of schematic flow chart of LBP feature extraction based on facial key points according to the embodiment of the present invention;

Fig. 4 is a schematic diagram of distribution of facial key points according to an embodiment of the present invention;

Fig. 5 is a schematic diagram of a face image to be processed before image correction according to an embodiment of the present invention;

Fig. 6 is a schematic diagram of a face image to be processed after image correction according to an embodiment of the present invention

Fig. 7 is a schematic diagram of the distribution of facial features according to an embodiment of the present invention;

Fig. 8 is a schematic diagram of distribution of facial key points for face pose estimation according to an embodiment of the present invention;

FIG. 9 is a schematic diagram of an ellipse neighborhood correction according to an embodiment of the present invention;

Fig. 10 is a schematic diagram of a face area division according to an embodiment of the present invention;

11 is a schematic diagram of another LBP feature extraction process based on facial key points according to an embodiment of the present invention;

Fig. 12 is a schematic diagram of an LBP feature extraction device based on facial key points according to an embodiment of the present invention.

Detailed ways

In order to make the purpose, technical solutions and advantages of the present invention clearer, the present invention will be further described in detail below in conjunction with the accompanying drawings. Obviously, the described embodiments are only some of the embodiments of the present invention, rather than all of them. Based on the embodiments of the present invention, all other embodiments obtained by persons of ordinary skill in the art without making creative efforts belong to the protection scope of the present invention.

In describing the present invention, it is to be understood that the terms "center", "upper", "lower", "front", "rear", "left", "right", "vertical", "horizontal", The orientations or positional relationships indicated by "top", "bottom", "inner", "outer", etc. are based on the orientations or positional relationships shown in the drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying References to devices or elements must have a particular orientation, be constructed, and operate in a particular orientation and therefore should not be construed as limiting the invention.

The terms "first", "second", "third", "fourth", "fifth", "sixth", "seventh", "eighth" and "ninth" are used for descriptive purposes only, It should not be understood as indicating or implying relative importance or implying the number of technical features indicated. Thus, the characteristics of "first", "second", "third", "fourth", "fifth", "sixth", "seventh", "eighth" and "ninth" are defined One or more of these features may be included explicitly or implicitly. In the description of the present invention, unless otherwise specified, "plurality" means two or more.

Embodiments of the present invention can be applied to various smart terminals, such as smart TVs, mobile phones, smart cameras, monitoring equipment, etc.; Security equipment, surveillance equipment, access control equipment, smart TVs and smartphones, etc.

Embodiment one

Fig. 3 shows a schematic flow chart of the LBP feature extraction process based on facial key points provided by an embodiment of the present invention. As shown in Fig. 3, the LBP feature extraction process based on facial key points includes:

Step 100: Estimating the face pose of the face image based on the key points of the face image to be processed.

Specifically, in the process of executing step 100, firstly, the face image to be processed is preprocessed to obtain a grayscale image of the image to be processed, and the facial key point detection is performed on the grayscale image of the image to be processed to obtain the The position of the facial key points to be extracted LBP features in the grayscale image. It should be noted that the face key points determined in step 100 include face contour key points, eyebrow key points, eye key points, nose key points and mouth key points, which respectively represent the face contour in the face image to be processed. The position of the eyebrows, the position of the eyes, the position of the nose and the position of the mouth.

As an example, Fig. 4 shows a schematic diagram of the distribution of facial key points according to an embodiment of the present invention. As shown in Fig. 4, there are 83 determined facial key points, wherein, the position and the size of the human face contour are represented There are 19 key points of the face contour, 16 key points of eyebrows representing the position and size of eyebrows, 18 key points of eyes representing the position and size of eyes, and 12 key points of nose representing the position and size of nose There are 18 mouth key points representing mouth position and mouth size. Of course, this is only an example, and does not mean that the distribution positions and numbers of facial key points determined in step 100 of the embodiment of the present invention are limited thereto.

Further, it should be noted that after the facial key point positions of the facial images to be processed are determined in step 100, it is necessary to perform image correction on the entire facial images to be processed, so as to ensure that each facial image to be processed , the angle between the line connecting the two eyeballs and the horizontal line is the same. For example, by rotating the entire face image to be processed, it is possible to ensure that in all the face images to be processed, the ordinates of the key points representing the positions of the two eyeballs of the human face are the same, that is, to ensure that the eyeballs of the human eyes are on the same horizontal line.

For example, as shown in FIG. 5, before the face image to be processed is not subjected to image correction, there is a certain angle between the line between the key points representing the positions of the two eyeballs of the human face and the horizontal line; referring to FIG. 6 , after image correction is performed on the face image to be processed, the line between the key points representing the positions of the two eyeballs of the human face is on the same horizontal line, that is, the ordinates of the key points representing the positions of the two eyeballs of the human face are the same; refer to Figure 5 and As shown in FIG. 6 , image correction is performed on the face image to be processed, only in the plane coordinate system, the face image to be processed is rotated as a whole, and the posture and size of the face in the face image to be processed are not changed.

Further, with reference to a large amount of experience, face poses usually include five different poses: front face, left face, right face, head down, and head up. Among them, the left face and right face are different from the front face. The facial features of the facial image change greatly in the horizontal direction, but do not change in the vertical direction; compared with the front face, the facial features of the facial image to be processed do not change in the horizontal direction. Larger changes in the vertical direction; based on the above analysis, the execution process of step 100 includes:

Obtaining a first vertical distance and a second vertical distance in the vertical direction of the facial image, wherein the first vertical distance and the second vertical distance are distributed from top to bottom in the vertical direction;

If the ratio between the first vertical distance and the second vertical distance is greater than a first threshold, then the facial posture of the facial image is head down; or the first vertical distance and the second vertical distance If the ratio between the vertical distances is smaller than the first threshold, the facial posture of the facial image is head up.

Obtaining a first horizontal distance and a second horizontal distance in the horizontal direction of the facial image, wherein the first horizontal distance and the second horizontal distance are distributed from left to right along the horizontal direction;

If the ratio between the first horizontal distance and the second horizontal distance is greater than a second threshold, then the face posture of the facial image is the right head; or the first horizontal distance and the second horizontal distance If the ratio between the distances is smaller than the second threshold, the face pose of the face image is the left head.

Specifically, the first vertical distance between the first key point and the second key point and the second vertical distance between the second key point and the third key point are acquired, wherein the first key point , the second key point and the third key point are distributed vertically from top to bottom;

If the ratio between the first vertical distance and the second vertical distance is greater than a first threshold, then the facial posture of the facial image is head down; or the first vertical distance and the second vertical distance If the ratio between the vertical distances is smaller than the first threshold, the facial posture of the facial image is head up.

Obtain the first horizontal distance between the third key point and the fourth key point and the second horizontal distance between the fifth key point and the sixth key point, wherein the third key point, the fourth key point , the fifth key point and the sixth key point are distributed from left to right along the horizontal direction;

If the ratio between the first horizontal distance and the second horizontal distance is greater than a second threshold, then the face posture of the facial image is the right head; or the first horizontal distance and the second horizontal distance If the ratio between the distances is smaller than the second threshold, the face pose of the face image is the left head.

It should be noted that during the execution of step 100, there is no sequence between judging whether the face posture is head up or down and judging whether the face posture is left head or right head, and the order of the two can be interchanged. , can also be executed simultaneously. That is, the execution process of step 100 can also be as follows:

Obtaining a first horizontal distance and a second horizontal distance in the horizontal direction of the facial image, wherein the first horizontal distance and the second horizontal distance are distributed from left to right along the horizontal direction;

If the ratio between the first horizontal distance and the second horizontal distance is greater than a second threshold, then the face posture of the facial image is the right head; or the first horizontal distance and the second horizontal distance If the ratio between the distances is smaller than the second threshold, the face pose of the face image is the left head.

Obtaining a first vertical distance and a second vertical distance in the vertical direction of the facial image, wherein the first vertical distance and the second vertical distance are distributed from top to bottom in the vertical direction;

If the ratio between the first vertical distance and the second vertical distance is greater than a first threshold, then the facial posture of the facial image is head down; or the first vertical distance and the second vertical distance If the ratio between the vertical distances is smaller than the first threshold, the facial posture of the facial image is head up.

Further, it should be noted that the second vertical distance may be the vertical distance between the second key point and the third key point, or the vertical distance between the third key point and the ninth key point, wherein , the first key point, the second key point, the third key point and the ninth key point are distributed vertically from top to bottom.

Further, it should be noted that the fourth key point and the fifth key point may be the same key point in the face image, or may be different key points distributed along the horizontal direction on the face image.

Further, it should be noted that the first threshold and the second threshold respectively represent the ratio of the first vertical distance to the second vertical distance and the ratio of the first horizontal distance to the second horizontal distance in the facial image in the front face state, The first threshold and the second threshold may be a specific numerical value or a numerical range respectively.

The execution process of step 100 of the embodiment of the present invention will be described in detail below in conjunction with a specific embodiment. Of course, this is only an example and does not mean that the face pose estimation of the embodiment of the present invention is limited thereto.

Figure 7 exemplarily shows a schematic diagram of the distribution of facial features in a frontal posture. Referring to Figure 7, the vertical distance between the eyeball and the frontal base line and the vertical distance between the eyeball and the highest point of the head of the face They respectively account for half of the vertical distance of the face, that is, the line connecting the two eyeballs of the face is the center line of the face in the vertical direction, and the vertical distance between the nose base line and the frontal base line is the same as The vertical distance between the eyebrow lines accounts for one-third of the vertical distance of the face, that is, the vertical distance between the eyebrow line and the bottom line of the nose and the vertical distance between the bottom line of the nose and the bottom line of the forehead in a frontal posture. The distance ratio is 1:1. Further analysis found that when the face posture changes from frontal to upward, the part of the face between the brow line and the bottom line of the nose is farther away from the camera lens than the part of the face between the bottom line of the nose and the bottom line of the forehead. The vertical distance between the eyebrow line and the bottom line of the nose is larger than that of the frontal face, and the vertical distance between the bottom line of the nose and the bottom line of the forehead is smaller than that of the frontal face. The ratio of the vertical distance between the brow line and the vertical distance between the nose base line and the forehead line is greater than 1:1; The face part between the forehead line and the base line is closer to the camera, the corresponding vertical distance between the brow line and the nose base line is smaller than that of the frontal face, and the vertical distance between the nose base line and the forehead line is relatively positive. When the face becomes larger, that is, when the head is lowered, the ratio of the vertical distance from the eyebrow line to the nose line to the vertical distance from the nose line to the forehead line is less than 1:1.

Based on the above analysis, the present invention can determine the face image to be processed by judging the ratio of the vertical distance between the eyebrow line to the nose base line and the vertical distance between the nose base line to the frontal base line and the first threshold Y Whether the face posture of the user is head down or head up. Considering the differences in measurement errors and the distribution of facial features of different faces, preferably, the value of the first threshold Y is a numerical range rather than a specific numerical value. For example, the first threshold Y is [0.8, 1.2].

Referring to Figure 8, the first key point A1 represents the vertical position of the eyebrow line in the facial image, the second key point A2 represents the vertical position of the nasal bottom line in the facial image, and the third key point A3 represents the forehead line In the vertical position in the facial image, the vertical distance Y1 between the first key point A1 and the second key point A2 represents the vertical distance between the eyebrow line and the bottom line of the nose, and the second key point A2 to the third key point The vertical distance between points A3 represents the vertical distance Y2 between the nasal base line and the frontal base line.

If the ratio of Y1 to Y2 is equal to the first threshold Y, that is, the ratio of Y1 to Y2 is within [0.8, 1.2], then the facial posture of the facial image is a positive face; if the ratio of Y1 to Y2 is greater than the first threshold Y , that is, the ratio of Y1 to Y2 is greater than 1.2, then the facial posture of the face image is head-up; if the ratio of Y1 to Y2 is less than the first threshold Y, that is, the ratio of Y1 to Y2 is less than 0.8, then the face image of the face The facial posture is head down.

Referring to Figure 7, among facial features in a frontal posture, the horizontal distance between the left corner of the left eye and the root of the left ear, the horizontal distance between the left corner of the left eye and the right corner of the left eye, the right corner of the left eye The horizontal distance between the corner of the eye and the left corner of the right eye, the horizontal distance between the left corner of the right eye and the right corner of the right eye, and the horizontal distance between the right corner of the right eye and the root of the right ear respectively account for the overall facial image One-fifth of the horizontal distance, that is, the horizontal distance between the left corner of the left eye and the base of the left ear, the horizontal distance between the left corner of the left eye and the right corner of the left eye, and the horizontal distance between the right corner of the left eye and the left corner of the right eye. The ratio of the horizontal distance between the corners of the eyes, the horizontal distance between the left corner of the right eye to the right corner of the right eye, and the horizontal distance between the right corner of the right eye and the root of the right ear is 1:1:1 :1:1:1. After research, it is found that when the face posture of the face image changes from the front face to the left face or the right face, the horizontal distance between the left corner of the left eye and the root of the left ear, the left corner of the left eye to the right of the left eye The horizontal distance between the corners of the eyes, the horizontal distance between the right corner of the left eye and the left corner of the right eye, the horizontal distance between the left corner of the right eye and the right corner of the right eye, and the distance between the right corner of the right eye and the root of the right ear The horizontal distances between the above five horizontal distances have different degrees of change relative to the face posture. Therefore, we can judge whether the face posture of the face image is the left side by judging the ratio change of any two of the above five horizontal distances. face or the right side of the face.

The ratio of the horizontal distance between the left corner of the left eye to the right corner of the left eye and the horizontal distance between the left corner of the right eye to the right corner of the right eye and the second threshold X will be combined to judge Whether the face pose of the image to be processed is left head or right head. Considering the differences in measurement errors and facial features, preferably, the value of the second threshold X is a numerical range rather than a specific numerical value. For example, the first threshold X is [0.9, 1.1]. Of course, this is just an example, and it does not mean that the method of judging whether the facial posture of the image to be processed is a left head or a right head is not limited in this embodiment of the present invention.

Specifically, as shown in FIG. 8 , the fourth key point B1 represents the horizontal position of the left corner of the left eye in the face image, the fifth key point B2 represents the horizontal position of the right corner of the left eye in the face image, and the fifth key point B2 represents the horizontal position of the right corner of the left eye in the face image. The six key points B3 represent the horizontal position of the left corner of the right eye in the face image, the seventh key point B4 represents the horizontal position of the right eye corner of the right eye in the face image, the fourth key point B1 to the fifth key point B2 The horizontal distance between represents the horizontal distance between the left corner of the left eye and the right corner of the left eye, and the horizontal distance X2 between the sixth key point B3 and the seventh key point B4 represents the distance from the left corner of the right eye to the right eye. The horizontal distance between the corners of the right eye.

If the ratio of X1 to X2 is equal to the second threshold X, that is, the ratio of X1 to X2 is within [0.9, 1.1], then the face posture of the facial image is a positive face; if the ratio of X1 to X2 is greater than the second threshold X , that is, the ratio of X1 to X2 is greater than 1.1, then the face pose of the face image is the left head; if the ratio of X1 to X2 is less than the second threshold X, that is, the ratio of X1 to X2 is less than 0.9, the face image The face pose of is right head.

Further, if the ratio of Y1 to Y2 is greater than the first threshold Y and the ratio of X1 to X2 is greater than the second threshold X, the facial posture of the face image is head up and left head; if the ratio of Y1 to Y2 is greater than The first threshold Y and the ratio of X1 to X2 is less than the second threshold X, then the facial posture of the facial image is head up and right head; if the ratio of Y1 to Y2 is less than the first threshold Y and the ratio of X1 to X2 If it is greater than the second threshold X, the facial posture of the facial image is head down and left head. If the ratio of Y1 to Y2 is smaller than the first threshold Y and the ratio of X1 to X2 is smaller than the second threshold X, the facial image The face posture of is head down and head on the right side.

Step 110: Correct the radius of the ellipse neighborhood corresponding to the key point according to the face pose of the face image.

Specifically, the execution process of step 110 is as follows:

If the face pose is a front face, the horizontal axis of the ellipse neighborhood is equal to a first preset value, and the vertical axis of the ellipse neighborhood is equal to a first preset value.

If the facial posture is head up or head down, the horizontal axis of the ellipse neighborhood is equal to a first preset value, and the vertical axis of the ellipse neighborhood is equal to the product of the first preset value and a first preset weight.

If the face posture is the left head or the right head, the horizontal axis of the ellipse neighborhood is equal to the product of the first preset value and the second preset weight, and the vertical axis of the ellipse neighborhood is equal to the first preset weight. set value.

Specifically, as shown in FIG. 9 , the horizontal axis of the elliptical neighborhood of the key point O is W, and the vertical axis is H. According to the judgment result of step 100, if the face pose of the face image to be processed is a frontal face, then the horizontal axis W of the ellipse neighborhood of the key point O is equal to the first preset value D, and the vertical axis H is equal to the first preset value D ; If the face posture of the face image to be processed is head up or head down, the horizontal axis W of the ellipse neighborhood of the key point O is equal to the first preset value D, and the vertical axis H is equal to the first preset value D and the first preset value. Set the product of weight &1; if the face pose of the face image to be processed is the left head or the right head, then the horizontal axis W of the ellipse neighborhood of the key point O is equal to the first preset value D and the second preset weight &2 The product of , the vertical axis H is equal to the first preset value D.

Further, firstly, the face image to be processed is divided into four regions, which are respectively the first region, the second region, the third region and the fourth region, wherein the first region, the second region, the third region and the fourth region The four areas are distributed clockwise, and the first area is located in the vertical direction of the fourth area.

For example, as shown in Figure 10, the first area is located on the left half of the face and above the nose line, the second area is located on the right half of the face and above the nose line, and the third area is located on the bottom of the face. The right half of the face is located in the area below the nasal base line, and the fourth area is located in the left half of the face and is located in the area below the nasal base line.

Specifically, according to the face pose estimation result of the face image to be processed in step 100, if the face pose of the face image to be processed is a frontal face, it is located in the first area, the second area, the third area and the fourth area The horizontal axis W of the ellipse neighborhood of the key point O is equal to the first preset value D, and the vertical axis H is equal to the first preset value D.

If the face posture of the face image to be processed is head-up, then the horizontal axis W of the ellipse neighborhood of the key point O located in the first area, the second area, the third area and the fourth area is equal to the first preset value D, The vertical axis H is equal to the product of the first preset value D and the first preset weight &1. Among them, when the key point O is located in the first area and the second area, the first preset weight &1 is less than 1, that is, the vertical axis H of the ellipse neighborhood of the key point O located in the first area and the second area is smaller than the first preset weight Set the value D; when the key point O is located in the third area and the fourth area, the first preset weight &1 is greater than 1, that is, the vertical axis H of the ellipse neighborhood of the key point O located in the first area and the second area is greater than the first A default value D.

If the face posture of the face image to be processed is head down, the horizontal axis W of the ellipse neighborhood of the key point O located in the first area, the second area, the third area and the fourth area is equal to the first preset value D, and the vertical axis is equal to the first preset value D. The axis H is equal to the product of the first preset value D and the first preset weight &1. Among them, when the key point O is located in the first area and the second area, the first preset weight &1 is greater than 1, that is, the vertical axis H of the ellipse neighborhood of the key point O located in the first area and the second area is greater than the first preset weight Set the value D; when the key point O is located in the third area and the fourth area, the first preset weight &1 is less than 1, that is, the vertical axis H of the ellipse neighborhood of the key point O located in the first area and the second area is smaller than the first A default value D.

If the face pose of the face image to be processed is the left head, the horizontal axis W of the ellipse neighborhood of the key point O located in the first area, the second area, the third area and the fourth area is equal to the first preset value D The product of the second preset weight &2, the vertical axis H is equal to the first preset value D. Among them, when the key point O is located in the first area and the third area, the second preset weight &2 is greater than 1, that is, the horizontal axis W of the ellipse neighborhood of the key point O located in the first area and the second area is greater than the first preset weight Set the value D; when the key point O is located in the second area and the fourth area, the second preset weight &2 is less than 1, that is, the horizontal axis W of the ellipse neighborhood of the key point O located in the first area and the second area is smaller than the first area A default value D.

If the face pose of the face image to be processed is the right head, the horizontal axis W of the ellipse neighborhood of the key point O located in the first area, the second area, the third area and the fourth area is equal to the first preset value D The product of the second preset weight &2, the vertical axis H is equal to the first preset value D. Among them, when the key point O is located in the first area and the third area, the second preset weight &2 is less than 1, that is, the horizontal axis W of the ellipse neighborhood of the key point O located in the first area and the second area is smaller than the first preset weight Set the value D; when the key point O is located in the second area and the fourth area, the second preset weight &2 is greater than 1, that is, the horizontal axis W of the ellipse neighborhood of the key point O located in the first area and the second area is greater than the first area A default value D.

If the face posture of the face image to be processed is head up and head on the left side, then the horizontal axis W of the ellipse neighborhood of the key point O located in the first area, the second area, the third area and the fourth area is equal to the first preset The product of the set value D and the second preset weight &2, and the vertical axis H are equal to the product of the first preset value D and the first preset weight &1. Wherein, when the key point O is located in the first area, the second preset weight &2 is greater than 1, that is, the horizontal axis W of the ellipse neighborhood of the key point O located in the first area is greater than the first preset value D, and the first preset The weight &1 is less than 1, that is, the vertical axis H of the ellipse neighborhood of the key point O located in the first area is smaller than the first preset value D; when the key point O is located in the second area, the second preset weight &2 is less than 1, that is The horizontal axis W of the ellipse neighborhood of the key point O located in the first area is less than the first preset value D, and the first preset weight &1 is less than 1, that is, the vertical axis H of the ellipse neighborhood of the key point O located in the first area Less than the first preset value D; when the key point O is located in the third area, the second preset weight &2 is greater than 1, that is, the horizontal axis W of the ellipse neighborhood of the key point O located in the first area is greater than the first preset value D, the first preset weight &1 is greater than 1, that is, the vertical axis H of the ellipse neighborhood of the key point O located in the first area is greater than the first preset value D; when the key point O is located in the fourth area, the second preset Weight &2 is less than 1, that is, the horizontal axis W of the ellipse neighborhood of the key point O in the first area is smaller than the first preset value D, and the first preset weight &1 is greater than 1, that is, the ellipse of the key point O in the first area The vertical axis H of the neighborhood is larger than the first preset value D.

If the facial posture of the face image to be processed is head up and head on the right side, then the horizontal axis W of the ellipse neighborhood of the key point O located in the first area, the second area, the third area and the fourth area is equal to the first preset The product of the set value D and the second preset weight &2, and the vertical axis H are equal to the product of the first preset value D and the first preset weight &1. Wherein, when the key point O is located in the first area, the second preset weight &2 is less than 1, that is, the horizontal axis W of the ellipse neighborhood of the key point O located in the first area is smaller than the first preset value D, and the first preset The weight &1 is less than 1, that is, the vertical axis H of the ellipse neighborhood of the key point O located in the first area is smaller than the first preset value D; when the key point O is located in the second area, the second preset weight &2 is greater than 1, that is The horizontal axis W of the ellipse neighborhood of the key point O located in the first area is greater than the first preset value D, and the first preset weight &1 is less than 1, that is, the vertical axis H of the ellipse neighborhood of the key point O located in the first area Less than the first preset value D; when the key point O is located in the third area, the second preset weight &2 is less than 1, that is, the horizontal axis W of the ellipse neighborhood of the key point O located in the first area is smaller than the first preset value D, the first preset weight &1 is greater than 1, that is, the vertical axis H of the ellipse neighborhood of the key point O located in the first area is greater than the first preset value D; when the key point O is located in the fourth area, the second preset Weight &2 is greater than 1, that is, the horizontal axis W of the ellipse neighborhood of the key point O in the first area is greater than the first preset value D, and the first preset weight &1 is greater than 1, that is, the ellipse of the key point O in the first area The vertical axis H of the neighborhood is larger than the first preset value D.

If the face posture of the face image to be processed is head down and head on the left side, the horizontal axis W of the ellipse neighborhood of the key point O located in the first area, the second area, the third area and the fourth area is equal to the first preset The product of the value D and the second preset weight &2, the vertical axis H is equal to the product of the first preset value D and the first preset weight &1. Wherein, when the key point O is located in the first area, the second preset weight &2 is greater than 1, that is, the horizontal axis W of the ellipse neighborhood of the key point O located in the first area is greater than the first preset value D, and the first preset The weight &1 is greater than 1, that is, the vertical axis H of the ellipse neighborhood of the key point O located in the first area is greater than the first preset value D; when the key point O is located in the second area, the second preset weight &2 is less than 1, that is The horizontal axis W of the ellipse neighborhood of the key point O located in the first area is less than the first preset value D, and the first preset weight &1 is greater than 1, that is, the vertical axis H of the ellipse neighborhood of the key point O located in the first area Greater than the first preset value D; when the key point O is located in the third area, the second preset weight &2 is greater than 1, that is, the horizontal axis W of the ellipse neighborhood of the key point O located in the first area is greater than the first preset value D, the first preset weight &1 is less than 1, that is, the vertical axis H of the ellipse neighborhood of the key point O located in the first area is smaller than the first preset value D; when the key point O is located in the fourth area, the second preset Weight &2 is less than 1, that is, the horizontal axis W of the ellipse neighborhood of the key point O in the first area is smaller than the first preset value D, and the first preset weight &1 is less than 1, that is, the ellipse of the key point O in the first area The vertical axis H of the neighborhood is smaller than the first preset value D.

If the face posture of the face image to be processed is head down and head on the right side, then the horizontal axis W of the ellipse neighborhood of the key point O located in the first area, the second area, the third area and the fourth area is equal to the first preset The product of the set value D and the second preset weight &2, and the vertical axis H are equal to the product of the first preset value D and the first preset weight &1. Wherein, when the key point O is located in the first area, the second preset weight &2 is less than 1, that is, the horizontal axis W of the ellipse neighborhood of the key point O located in the first area is smaller than the first preset value D, and the first preset The weight &1 is greater than 1, that is, the vertical axis H of the ellipse neighborhood of the key point O located in the first area is greater than the first preset value D; when the key point O is located in the second area, the second preset weight &2 is greater than 1, that is The horizontal axis W of the ellipse neighborhood of the key point O located in the first area is greater than the first preset value D, and the first preset weight &1 is greater than 1, that is, the vertical axis H of the ellipse neighborhood of the key point O located in the first area Greater than the first preset value D; when the key point O is located in the third area, the second preset weight &2 is less than 1, that is, the horizontal axis W of the ellipse neighborhood of the key point O located in the first area is smaller than the first preset value D, the first preset weight &1 is less than 1, that is, the vertical axis H of the ellipse neighborhood of the key point O located in the first area is smaller than the first preset value D; when the key point O is located in the fourth area, the second preset Weight &2 is greater than 1, that is, the horizontal axis W of the ellipse neighborhood of the key point O in the first area is greater than the first preset value D, and the first preset weight &1 is less than 1, that is, the ellipse of the key point O in the first area The vertical axis H of the neighborhood is smaller than the first preset value D.

For example, when the first preset weight &1 is less than 1, &1=1-e·sinα, wherein, α is the side head angle of the face posture in the face image to be processed; when the first preset weight &1 is greater than 1, &1 =1+e sin α, wherein, α is the side head angle of face attitude in the face image to be processed; When the second preset weight & 2 is less than 1, & 2=0.8 of example; When the second preset weight & 2 is greater than 1 , &2=1.2 for the example.

For example, in the face image of the same person, the corresponding relationship between the horizontal axis W and the vertical axis H of the ellipse neighborhood of the key point O, the face pose and the position of the key point O is shown in Table 1 below:

Table I

Step 120: performing LBP feature extraction on the key points using the ellipse neighborhood radius on the face image.

Specifically, before performing LBP feature extraction on the key points of the face image to be processed, it is necessary to perform normalization processing on the face image to be processed, and scale the size of the face image to be processed to the same size. For example, if the actual distance between the two eyeballs of the face in the face image to be processed is L, if the face pose of the face image to be processed is not the left head or the right head, then the face image to be processed is scaled until it The distance between the two eyeballs of the face is L; if the face pose of the face image to be processed is the left head or the right head, then the face image to be processed is scaled until the distance between the two eyeballs of the face is L cosα , where α is the side head angle of the face pose in the face image to be processed.

Further, take each key point as the center, take an area in the face image to be processed, and divide the area into several image blocks of the same size, and then perform LBP feature extraction on each pixel of each image block . In the process of LBP feature extraction for each pixel point, a circular LBP operator is used, wherein the neighborhood radius of each pixel point adopts the transverse radius of the ellipse neighborhood determined according to the face pose and key point position in step 110. Axis W and vertical axis H, and then obtain the pixel values of several sampling points in the ellipse neighborhood, compare the pixel values of the several sampling points with the pixel value of the center point of the ellipse neighborhood, and the pixels greater than the center point The sampling point position of the value is marked as 1, otherwise it is marked as 0. Then, according to the marking results of each sampling point, a binary number with a certain number of digits is obtained in a certain order, and then the binary number is converted into a decimal number according to the equivalent mode, and the histogram of the image block is generated according to the decimal number, and finally The histograms of each image block are concatenated to generate the LBP feature vector of the key point.

Further, the present invention does not introduce in detail about the LBP feature extraction algorithm of key points, and those skilled in the art may refer to the prior art.

The face key point-based LBP feature extraction method provided by the embodiment of the present invention first determines the face pose of the face image according to the key points of the face image, and then corrects different face key points according to the face pose of the face image. The corresponding relationship between the radius of the ellipse neighborhood and the face pose of the face image and the position of the key point is established, and the corrected radius of the ellipse neighborhood is used to extract the key points of the face. Point LBP features, thus avoiding the problem that the local scaling ratio of the face image is different under different face poses of the same person, resulting in a large difference in the LBP features extracted from the same key point, and improving the face key points based on different face poses. The effectiveness of the LBP feature extraction, thereby improving the accuracy of face recognition.

Embodiment two

Fig. 11 shows a schematic diagram of another LBP feature extraction process based on facial key points provided by an embodiment of the present invention. As shown in Fig. 11, the LBP feature extraction process based on facial key points includes:

Step 200: Estimating the face pose of the face image based on the key points of the face image to be processed.

Specifically, the execution process of step 200 is the same as that of step 100 in the above-mentioned embodiment 1, which has been introduced in detail above, and the present invention will not be repeated here. For details, please refer to step 100 in embodiment 1.

Step 210: Determine the radius of the ellipse neighborhood corresponding to the key point in the preset mapping table according to the face pose of the face image.

Specifically, according to the face pose of the face image to be processed determined in the above step 200, the ellipse neighborhood radius corresponding to the key point of the face image to be processed in the preset mapping table is searched, wherein the preset mapping table It is used to represent the face image of the same person, and the corresponding ellipse neighborhood radius in the process of LBP feature extraction of key points in different face regions under different face poses. Exemplarily, as shown in FIG. 10 , the face is divided into four regions, namely the first region, the second region, the third region and the fourth region, wherein the first region, the second region, the third region and the The fourth area is distributed clockwise, and the first area is located in the vertical direction of the fourth area, wherein the first area is located on the left half of the face and above the nose line, and the second area is located on the right side of the face. The half face is located above the nose line, the third area is located on the right half of the face and below the nose line, and the fourth area is located on the left half of the face and below the nose line.

Further, in the preset mapping table, the corresponding relationship between the key point O of the face image to be processed and the face pose of the face image to be processed and the distribution position of the key point O on the face image to be processed is as follows:

If the face pose of the face image to be processed is a positive face, then the horizontal axis W of the ellipse neighborhood of the key point O located in the first area, the second area, the third area and the fourth area is equal to the first preset value D, The vertical axis H is equal to the first preset value D.

If the face posture of the face image to be processed is head-up, then the horizontal axis W of the ellipse neighborhood of the key point O located in the first area, the second area, the third area and the fourth area is equal to the first preset value D, The vertical axis H is equal to the product of the first preset value D and the first preset weight &1. Among them, when the key point O is located in the first area and the second area, the first preset weight &1 is less than 1, that is, the vertical axis H of the ellipse neighborhood of the key point O located in the first area and the second area is smaller than the first preset weight Set the value D; when the key point O is located in the third area and the fourth area, the first preset weight &1 is greater than 1, that is, the vertical axis H of the ellipse neighborhood of the key point O located in the first area and the second area is greater than the first A default value D.

If the face posture of the face image to be processed is head down, the horizontal axis W of the ellipse neighborhood of the key point O located in the first area, the second area, the third area and the fourth area is equal to the first preset value D, and the vertical axis is equal to the first preset value D. The axis H is equal to the product of the first preset value D and the first preset weight &1. Among them, when the key point O is located in the first area and the second area, the first preset weight &1 is greater than 1, that is, the vertical axis H of the ellipse neighborhood of the key point O located in the first area and the second area is greater than the first preset weight Set the value D; when the key point O is located in the third area and the fourth area, the first preset weight &1 is less than 1, that is, the vertical axis H of the ellipse neighborhood of the key point O located in the first area and the second area is smaller than the first A default value D.

If the face pose of the face image to be processed is the left head, the horizontal axis W of the ellipse neighborhood of the key point O located in the first area, the second area, the third area and the fourth area is equal to the first preset value D The product of the second preset weight &2, the vertical axis H is equal to the first preset value D. Among them, when the key point O is located in the first area and the third area, the second preset weight &2 is greater than 1, that is, the horizontal axis W of the ellipse neighborhood of the key point O located in the first area and the second area is greater than the first preset weight Set the value D; when the key point O is located in the second area and the fourth area, the second preset weight &2 is less than 1, that is, the horizontal axis W of the ellipse neighborhood of the key point O located in the first area and the second area is smaller than the first area A default value D.

If the face pose of the face image to be processed is the right head, the horizontal axis W of the ellipse neighborhood of the key point O located in the first area, the second area, the third area and the fourth area is equal to the first preset value D The product of the second preset weight &2, the vertical axis H is equal to the first preset value D. Among them, when the key point O is located in the first area and the third area, the second preset weight &2 is less than 1, that is, the horizontal axis W of the ellipse neighborhood of the key point O located in the first area and the second area is smaller than the first preset weight Set the value D; when the key point O is located in the second area and the fourth area, the second preset weight &2 is greater than 1, that is, the horizontal axis W of the ellipse neighborhood of the key point O located in the first area and the second area is greater than the first area A default value D.

If the face posture of the face image to be processed is head up and head on the left side, then the horizontal axis W of the ellipse neighborhood of the key point O located in the first area, the second area, the third area and the fourth area is equal to the first preset The product of the set value D and the second preset weight &2, and the vertical axis H are equal to the product of the first preset value D and the first preset weight &1. Wherein, when the key point O is located in the first area, the second preset weight &2 is greater than 1, that is, the horizontal axis W of the ellipse neighborhood of the key point O located in the first area is greater than the first preset value D, and the first preset The weight &1 is less than 1, that is, the vertical axis H of the ellipse neighborhood of the key point O located in the first area is smaller than the first preset value D; when the key point O is located in the second area, the second preset weight &2 is less than 1, that is The horizontal axis W of the ellipse neighborhood of the key point O located in the first area is less than the first preset value D, and the first preset weight &1 is less than 1, that is, the vertical axis H of the ellipse neighborhood of the key point O located in the first area Less than the first preset value D; when the key point O is located in the third area, the second preset weight &2 is greater than 1, that is, the horizontal axis W of the ellipse neighborhood of the key point O located in the first area is greater than the first preset value D, the first preset weight &1 is greater than 1, that is, the vertical axis H of the ellipse neighborhood of the key point O located in the first area is greater than the first preset value D; when the key point O is located in the fourth area, the second preset Weight &2 is less than 1, that is, the horizontal axis W of the ellipse neighborhood of the key point O in the first area is smaller than the first preset value D, and the first preset weight &1 is greater than 1, that is, the ellipse of the key point O in the first area The vertical axis H of the neighborhood is larger than the first preset value D.

If the facial posture of the face image to be processed is head up and head on the right side, then the horizontal axis W of the ellipse neighborhood of the key point O located in the first area, the second area, the third area and the fourth area is equal to the first preset The product of the set value D and the second preset weight &2, and the vertical axis H are equal to the product of the first preset value D and the first preset weight &1. Wherein, when the key point O is located in the first area, the second preset weight &2 is less than 1, that is, the horizontal axis W of the ellipse neighborhood of the key point O located in the first area is smaller than the first preset value D, and the first preset The weight &1 is less than 1, that is, the vertical axis H of the ellipse neighborhood of the key point O located in the first area is smaller than the first preset value D; when the key point O is located in the second area, the second preset weight &2 is greater than 1, that is The horizontal axis W of the ellipse neighborhood of the key point O located in the first area is greater than the first preset value D, and the first preset weight &1 is less than 1, that is, the vertical axis H of the ellipse neighborhood of the key point O located in the first area Less than the first preset value D; when the key point O is located in the third area, the second preset weight &2 is less than 1, that is, the horizontal axis W of the ellipse neighborhood of the key point O located in the first area is smaller than the first preset value D, the first preset weight &1 is greater than 1, that is, the vertical axis H of the ellipse neighborhood of the key point O located in the first area is greater than the first preset value D; when the key point O is located in the fourth area, the second preset Weight &2 is greater than 1, that is, the horizontal axis W of the ellipse neighborhood of the key point O in the first area is greater than the first preset value D, and the first preset weight &1 is greater than 1, that is, the ellipse of the key point O in the first area The vertical axis H of the neighborhood is larger than the first preset value D.

If the face posture of the face image to be processed is head down and head on the left side, the horizontal axis W of the ellipse neighborhood of the key point O located in the first area, the second area, the third area and the fourth area is equal to the first preset The product of the value D and the second preset weight &2, the vertical axis H is equal to the product of the first preset value D and the first preset weight &1. Wherein, when the key point O is located in the first area, the second preset weight &2 is greater than 1, that is, the horizontal axis W of the ellipse neighborhood of the key point O located in the first area is greater than the first preset value D, and the first preset The weight &1 is greater than 1, that is, the vertical axis H of the ellipse neighborhood of the key point O located in the first area is greater than the first preset value D; when the key point O is located in the second area, the second preset weight &2 is less than 1, that is The horizontal axis W of the ellipse neighborhood of the key point O located in the first area is less than the first preset value D, and the first preset weight &1 is greater than 1, that is, the vertical axis H of the ellipse neighborhood of the key point O located in the first area Greater than the first preset value D; when the key point O is located in the third area, the second preset weight &2 is greater than 1, that is, the horizontal axis W of the ellipse neighborhood of the key point O located in the first area is greater than the first preset value D, the first preset weight &1 is less than 1, that is, the vertical axis H of the ellipse neighborhood of the key point O located in the first area is smaller than the first preset value D; when the key point O is located in the fourth area, the second preset Weight &2 is less than 1, that is, the horizontal axis W of the ellipse neighborhood of the key point O in the first area is smaller than the first preset value D, and the first preset weight &1 is less than 1, that is, the ellipse of the key point O in the first area The vertical axis H of the neighborhood is smaller than the first preset value D.

If the face posture of the face image to be processed is head down and head on the right side, then the horizontal axis W of the ellipse neighborhood of the key point O located in the first area, the second area, the third area and the fourth area is equal to the first preset The product of the set value D and the second preset weight &2, and the vertical axis H are equal to the product of the first preset value D and the first preset weight &1. Wherein, when the key point O is located in the first area, the second preset weight &2 is less than 1, that is, the horizontal axis W of the ellipse neighborhood of the key point O located in the first area is smaller than the first preset value D, and the first preset The weight &1 is greater than 1, that is, the vertical axis H of the ellipse neighborhood of the key point O located in the first area is greater than the first preset value D; when the key point O is located in the second area, the second preset weight &2 is greater than 1, that is The horizontal axis W of the ellipse neighborhood of the key point O located in the first area is greater than the first preset value D, and the first preset weight &1 is greater than 1, that is, the vertical axis H of the ellipse neighborhood of the key point O located in the first area Greater than the first preset value D; when the key point O is located in the third area, the second preset weight &2 is less than 1, that is, the horizontal axis W of the ellipse neighborhood of the key point O located in the first area is smaller than the first preset value D, the first preset weight &1 is less than 1, that is, the vertical axis H of the ellipse neighborhood of the key point O located in the first area is smaller than the first preset value D; when the key point O is located in the fourth area, the second preset Weight &2 is greater than 1, that is, the horizontal axis W of the ellipse neighborhood of the key point O in the first area is greater than the first preset value D, and the first preset weight &1 is less than 1, that is, the ellipse of the key point O in the first area The vertical axis H of the neighborhood is smaller than the first preset value D.

For example, when the first preset weight &1 is less than 1, &1=1-e·sinα, wherein, α is the side head angle of the face posture in the face image to be processed; when the first preset weight &1 is greater than 1, &1 =1+e sin α, wherein, α is the side head angle of face attitude in the face image to be processed; When the second preset weight & 2 is less than 1, & 2=0.7 of example; When the second preset weight & 2 is greater than 1 , &2=1.1 for the example.

For example, for the face image of the same person, the corresponding relationship between the horizontal axis W and the vertical axis H of the ellipse neighborhood of the key point O, the face posture and the position of the key point O is shown in Table 2 below, that is, the preset The mapping table is shown in Table 2 below:

Table II

Step 220: Perform LBP feature extraction on the key points using the ellipse neighborhood radius on the face image.

Specifically, the execution process of step 220 is the same as that of step 120 in the above-mentioned embodiment 1, which has been introduced in detail above, and the present invention will not be repeated here. For details, please refer to step 120 of embodiment 1.

The LBP feature extraction method based on facial key points provided by the embodiment of the present invention first determines the face pose of the face image according to the key points of the face image, and then searches for the face to be processed according to the face pose of the face image. The key points of the internal image correspond to the ellipse neighborhood radius in the preset mapping table, where the preset mapping table is used to represent the face image of the same person, and the key points in different face regions under different face poses The radius of the corresponding ellipse neighborhood during the LBP feature extraction process of the point. The preset mapping table in the embodiment of the present invention establishes the corresponding relationship between the radius of the ellipse neighborhood and the face pose of the facial image and the position of the key point, and then uses the radius of the neighborhood of the ellipse in the preset mapping table to extract the The LBP feature of the key points of the face, thus avoiding the problem that the local scaling ratio of the face image is different under different face poses of the same person, resulting in a large difference in the LBP features extracted from the same key point, and improving the face based on different face poses. The effectiveness of the LBP feature extraction of internal key points, thereby improving the accuracy of face recognition.

Based on the LBP feature extraction method based on the facial key points of the above-mentioned embodiment, the embodiment of the present invention also provides an LBP feature extraction device based on the facial key points, and the LBP feature extraction method provided by the above-mentioned embodiment is used to extract the face to be processed. The LBP feature extraction of the image, the device is shown in Figure 12, including:

The key point detection module 301 is used to detect the key points of the face image to be processed, and determine the position of the key points of the face to be extracted from the LBP feature.

The image processing module 302 is configured to obtain the distance between the key points, determine the face pose of the face image according to the distance, and correct the radius of the ellipse neighborhood corresponding to the key point according to the face pose.

The feature extraction module 303 is configured to perform LBP feature extraction on the key points using the radius of the ellipse neighborhood on the face image.

Specifically, the key point detection module 301 is used to detect the key points of the face image to be processed, and obtain the position of the key points of the face to be extracted from the face image to be processed. Further, the key point detection module 301 is used to detect the key points of the face profile, the eyebrow key, the position of the eyebrow, the position of the eye, the position of the nose and the position of the mouth respectively representing the position of the face profile in the face image to be processed. Points, eye key points, nose key points and mouth key points, etc., and then based on the position of the key points, the face image to be processed is divided into the first area, the second area, the third area and the fourth area as shown in Figure 10 Four areas, wherein the first area, the second area, the third area and the fourth area are distributed clockwise, and the first area is located in the vertical direction of the fourth area, for example, the first area is located in the face The left half of the face is located above the nose line, the second area is located on the right half of the face and above the nose line, the third area is located on the right half of the face and below the nose line, and the fourth area is located on the The left half of the face and the area below the bottom line of the nose.

Specifically, the image processing module 302 is used to obtain the distance between the key points to be processed, and determine the face pose of the face image to be processed according to the obtained distance between the key points of the face, and finally according to the determined to-be-processed The face pose of the face image is scaled to the face image to be processed, and the key points in different face regions under different face poses are determined to perform the corresponding ellipse neighborhood radius in the process of LBP feature extraction. Specifically, the image processing module 302 determines the face pose of the face image to be processed according to the acquired distance between facial key points, and performs processing on the face image to be processed according to the determined face pose of the face image to be processed. Scaling, and determining the key points in different face areas under different face postures to perform the method of corresponding ellipse neighborhood radius in the process of LBP feature extraction, and the steps 100 and 110 of the previous embodiment 1 and the method of embodiment 2 The methods described in step 200 and step 210 are the same, and the present invention will not repeat them here.

Specifically, the feature extraction module 303 is used to perform LBP feature extraction on the key point using the ellipse neighborhood radius of the facial key point determined in the image processing module 302 on the scaled face image, and the specific LBP feature extraction method and process , please refer to step 120 of Embodiment 1 and step 220 of Embodiment 2, and the present invention will not repeat them here.

The LBP feature extraction device based on facial key points provided by the embodiment of the present invention first determines the face pose of the face image according to the key points of the face image, and then determines the face to be processed according to the face pose of the face image. The corresponding radius of the ellipse neighborhood during the LBP feature extraction process of key points in different face regions of the face image under the face pose. The LBP feature extraction device in the embodiment of the present invention establishes the corresponding relationship between the radius of the ellipse neighborhood and the face pose of the face image and the position of the key point, and then uses the radius of the ellipse neighborhood to extract the key point of the face. LBP features, thus avoiding the problem that the local scaling ratio of the face image is different under different face poses of the same person, resulting in a large difference in the LBP features extracted from the same key point, and improving the LBP based on face key points under different face poses The effectiveness of feature extraction improves the accuracy of face recognition.

The present invention is described with reference to flowchart illustrations and/or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It should be understood that each procedure and/or block in the flowchart and/or block diagram, and a combination of procedures and/or blocks in the flowchart and/or block diagram can be realized by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special purpose computer, an embedded processor, or other programmable data processing equipment, so that the instructions executed by the processor of the computer or other programmable data processing equipment can realize the A process or processes and/or a function specified in a block or blocks of a block diagram.

These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing apparatus to operate in a specific manner, such that the instructions stored in the computer-readable memory produce an article of manufacture comprising instruction means, the instructions The device realizes the function specified in one or more procedures of the flowchart and/or one or more blocks of the block diagram.

These computer program instructions can also be loaded onto a computer or other programmable data processing device, causing a series of operational steps to be performed on the computer or other programmable device to produce a computer-implemented process, thereby The instructions provide steps for implementing the functions specified in the flow or flows of the flowcharts and/or the block or blocks of the block diagrams.

While preferred embodiments of the present invention have been described, additional changes and modifications can be made to these embodiments by those skilled in the art once the basic inventive concept is appreciated. Therefore, it is intended that the appended claims be construed to cover the preferred embodiment as well as all changes and modifications which fall within the scope of the invention.

Apparently, those skilled in the art can make various changes and modifications to the embodiments of the present invention without departing from the spirit and scope of the embodiments of the present invention. In this way, if the modifications and variations of the embodiments of the present invention fall within the scope of the claims of the present invention and equivalent technologies, the present invention also intends to include these modifications and variations.

Claims (19)

1. a kind of LBP feature extracting method based on face's key point characterized by comprising

Based on the key point of face image to be processed, the human face posture of the face image is estimated；

According to the human face posture of the face image, the corresponding Ellipse Neighborhood radius of the key point is corrected；

LBP feature extraction is carried out to the key point using the Ellipse Neighborhood radius to the face image.

2. a kind of LBP feature extracting method based on face's key point characterized by comprising

Based on the key point of face image to be processed, the human face posture of the face image is estimated；

The key point corresponding Ellipse Neighborhood radius in preset mapping table is determined according to the human face posture of the face image；

LBP feature extraction is carried out to the key point using the Ellipse Neighborhood radius to the face image.

3. method according to claim 1 or 2, which is characterized in that the key point based on face image to be processed is estimated Count the human face posture of the face image, comprising:

Obtain the first vertical distance and the second vertical distance on the face image vertical direction, wherein described first is vertical Distance and second vertical distance are distributed from up to down along the vertical direction；

Ratio between first vertical distance and second vertical distance is greater than first threshold, then the face image Human face posture is to bow；Or

Ratio between first vertical distance and second vertical distance is less than first threshold, then the face image Human face posture is to face upward head.

4. method according to claim 1 or claim 2, which is characterized in that the key point based on face image to be processed, estimation The human face posture of the face image, further includes:

Obtain the first level distance and the second horizontal distance in the face image horizontal direction, wherein the first level Distance and second horizontal distance are distributed from left to right in the horizontal direction；

Ratio between the first level distance and second horizontal distance is greater than second threshold, then the face image Human face posture is right side head；Or

Ratio between the first level distance and second horizontal distance is less than second threshold, then the face image Human face posture is left side head.

5. according to claim 1 or any one of 2 the methods, which is characterized in that the face appearance according to the face image State corrects the corresponding Ellipse Neighborhood radius of the key point, comprising:

The horizontal axis of the face if the human face posture is positive, the Ellipse Neighborhood is equal to the first preset value, the longitudinal axis of the Ellipse Neighborhood Equal to the first preset value；Or

If the human face posture is to face upward head or bow, the horizontal axis of the Ellipse Neighborhood is equal to the first preset value, the Ellipse Neighborhood The longitudinal axis be equal to the first preset value and the first default weight product；Or

If the human face posture is left side head or right side head, the horizontal axis of the Ellipse Neighborhood, which is equal to the first preset value and second, to be preset The longitudinal axis of the product of weight, the Ellipse Neighborhood is equal to the first preset value.

6. method according to claim 5, which is characterized in that the face image includes first area, second area, third Region and the fourth region, wherein the first area, the second area, the third region and the fourth region are along suitable Hour hands distribution, and the first area is located on the fourth region vertical direction.

7. method according to claim 6, which is characterized in that the human face posture according to the face image corrects institute State the corresponding Ellipse Neighborhood radius of key point, comprising:

The face if human face posture is positive, the first area, the second area, the third region and the fourth region The horizontal axis of the Ellipse Neighborhood and the longitudinal axis of the Ellipse Neighborhood be equal to the first preset value.

8. method according to claim 6, which is characterized in that the human face posture according to the face image corrects institute State the corresponding Ellipse Neighborhood radius of key point, further includes:

If the human face posture is to face upward head, the first area, the second area, the third region and the fourth region The horizontal axis be equal to first preset value, the longitudinal axis of the first area and the second area is default less than first The longitudinal axis of value, the third region and the fourth region is greater than first preset value.

9. method according to claim 6, which is characterized in that the human face posture according to the face image corrects institute State the corresponding Ellipse Neighborhood radius of key point, further includes:

If the human face posture is to bow, the first area, the second area, the third region and the fourth region The horizontal axis be equal to first preset value, it is default that the longitudinal axis of the first area and the second area is greater than first The longitudinal axis of value, the third region and the fourth region is less than first preset value.

10. method according to claim 6, which is characterized in that the human face posture according to the face image corrects institute State the corresponding Ellipse Neighborhood radius of key point, further includes:

If the human face posture is left side head, it is default that the horizontal axis of the first area and the fourth region is greater than first Value, the horizontal axis in the second area and the third region are less than first preset value, the first area, described the The longitudinal axis in two regions, the third region and the fourth region is equal to first preset value.

11. method according to claim 6, which is characterized in that the human face posture according to the face image corrects institute State the corresponding Ellipse Neighborhood radius of key point, further includes:

If the human face posture is right side head, the horizontal axis of the first area and the fourth region is default less than first Value, the horizontal axis in the second area and the third region are greater than first preset value, the first area, described the The longitudinal axis in two regions, the third region and the fourth region is equal to first preset value.

12. method according to claim 6, which is characterized in that the human face posture according to the face image corrects institute State the corresponding Ellipse Neighborhood radius of key point, further includes:

If the human face posture is to face upward head and left side head, the horizontal axis of the first area and the fourth region is greater than described The horizontal axis in the first preset value, the second area and the third region is less than first preset value, firstth area The longitudinal axis of domain and the second area is less than the first preset value, the longitudinal axis in the third region and the fourth region Greater than the first preset value.

13. method according to claim 6, which is characterized in that the human face posture according to the face image corrects institute State the corresponding Ellipse Neighborhood radius of key point, further includes:

If the human face posture is to face upward head and right side head, the horizontal axis of the first area and the fourth region is less than described The horizontal axis in the first preset value, the second area and the third region is greater than first preset value, firstth area The longitudinal axis of domain and the second area is greater than the first preset value, the longitudinal axis in the third region and the fourth region Less than the first preset value.

14. method according to claim 6, which is characterized in that the human face posture according to the face image corrects institute State the corresponding Ellipse Neighborhood radius of key point, further includes:

If the human face posture is to bow and left side head, the horizontal axis of the first area and the fourth region is greater than described The horizontal axis in the first preset value, the second area and the third region is less than first preset value, firstth area The longitudinal axis of domain and the second area is greater than the first preset value, the longitudinal axis in the third region and the fourth region Less than the first preset value.

15. method according to claim 6, which is characterized in that the human face posture according to the face image corrects institute State the corresponding Ellipse Neighborhood radius of key point, further includes:

If the human face posture is to face upward head and left side head, the horizontal axis of the first area and the fourth region is less than described The horizontal axis in the first preset value, the second area and the third region is greater than first preset value, firstth area The longitudinal axis of domain and the second area is less than the first preset value, the longitudinal axis in the third region and the fourth region Greater than the first preset value.

16. method according to claim 6, which is characterized in that preset mapping table is used to characterize under different faces posture, is different The corresponding Ellipse Neighborhood radius of the key point in face area, in which:

The face if human face posture is positive, the first area, the second area, the third region and the fourth region The horizontal axis of the Ellipse Neighborhood and the longitudinal axis of the Ellipse Neighborhood be equal to the first preset value；

If the human face posture is to face upward head, the first area, the second area, the third region and the fourth region The horizontal axis be equal to first preset value, the longitudinal axis of the first area and the second area is default less than first The longitudinal axis of value, the third region and the fourth region is greater than first preset value；

If the human face posture is to bow, the first area, the second area, the third region and the fourth region The horizontal axis be equal to first preset value, it is default that the longitudinal axis of the first area and the second area is greater than first The longitudinal axis of value, the third region and the fourth region is less than first preset value；

If the human face posture is left side head, it is default that the horizontal axis of the first area and the fourth region is greater than first Value, the horizontal axis in the second area and the third region are less than first preset value, the first area, described the The longitudinal axis in two regions, the third region and the fourth region is equal to first preset value；

If the human face posture is right side head, the horizontal axis of the first area and the fourth region is default less than first Value, the horizontal axis in the second area and the third region are greater than first preset value, the first area, described the The longitudinal axis in two regions, the third region and the fourth region is equal to first preset value；

If the human face posture is to face upward head and left side head, the horizontal axis of the first area and the fourth region is greater than described The horizontal axis in the first preset value, the second area and the third region is less than first preset value, firstth area The longitudinal axis of domain and the second area is less than the first preset value, the longitudinal axis in the third region and the fourth region Greater than the first preset value；

If the human face posture is to face upward head and right side head, the horizontal axis of the first area and the fourth region is less than described The horizontal axis in the first preset value, the second area and the third region is greater than first preset value, firstth area The longitudinal axis of domain and the second area is greater than the first preset value, the longitudinal axis in the third region and the fourth region Less than the first preset value；

If the human face posture is to bow and left side head, the horizontal axis of the first area and the fourth region is greater than described The horizontal axis in the first preset value, the second area and the third region is less than first preset value, firstth area The longitudinal axis of domain and the second area is greater than the first preset value, the longitudinal axis in the third region and the fourth region Less than the first preset value；

If the human face posture is to face upward head and left side head, the horizontal axis of the first area and the fourth region is less than described The horizontal axis in the first preset value, the second area and the third region is greater than first preset value, firstth area The longitudinal axis of domain and the second area is less than the first preset value, the longitudinal axis in the third region and the fourth region Greater than the first preset value.

17. a kind of LBP feature deriving means based on face's key point characterized by comprising

Critical point detection module, for detecting the key point of face image to be processed；

Image processing module determines the face image according to the distance for obtaining the distance between described key point Human face posture, and the radius that the key point corresponds to Ellipse Neighborhood is corrected according to the human face posture；

Characteristic extracting module, for carrying out LBP to the key point using the radius of the Ellipse Neighborhood to the face image Feature extraction.

18. 7 described device according to claim 1, which is characterized in that described image processing module is specifically used for:

Obtain the first vertical distance and the second vertical distance on the face image vertical direction, wherein described first is vertical Distance and second vertical distance are distributed from up to down along the vertical direction, first vertical distance with described second vertically away from Ratio between is greater than first threshold, then the human face posture of the face image be bow or first vertical distance with Ratio between second vertical distance is less than first threshold, then the human face posture of the face image is to face upward head；

Obtain the first level distance and the second horizontal distance in the face image horizontal direction, wherein the first level Distance and second horizontal distance are distributed from left to right in the horizontal direction, the first level distance with described second it is horizontal away from Ratio between is greater than second threshold, then the human face posture of the face image is right side head or the first level distance Ratio between second horizontal distance is less than second threshold, then the human face posture of the face image is left side head.

19. 7 described device according to claim 1, which is characterized in that described image processing module is specifically used for:

The face if human face posture is positive determines the described oval adjacent of first area, second area, third region and the fourth region The longitudinal axis of the horizontal axis in domain and the Ellipse Neighborhood is equal to the first preset value；Or

If the human face posture is to face upward head, the first area, the second area, the third region and the described 4th are determined The horizontal axis in region is equal to first preset value, and the longitudinal axis of the first area and the second area is less than first The longitudinal axis of preset value, the third region and the fourth region is greater than first preset value；Or

If the human face posture is to bow, the first area, the second area, the third region and the described 4th are determined The horizontal axis in region is equal to first preset value, and the longitudinal axis of the first area and the second area is greater than first The longitudinal axis of preset value, the third region and the fourth region is less than first preset value；Or

If the human face posture is left side head, determine that the horizontal axis of the first area and the fourth region is greater than first in advance If value, the horizontal axis in the second area and the third region is less than first preset value, the first area, described The longitudinal axis of second area, the third region and the fourth region is equal to first preset value；Or

If the human face posture is right side head, determine that the horizontal axis of the first area and the fourth region is pre- less than first If value, the horizontal axis in the second area and the third region is greater than first preset value, the first area, described The longitudinal axis of second area, the third region and the fourth region is equal to first preset value；Or

If the human face posture is to face upward head and left side head, determine that the horizontal axis of the first area and the fourth region is greater than The horizontal axis in first preset value, the second area and the third region is less than first preset value, and described the The longitudinal axis of one region and the second area less than the first preset value, the third region and the fourth region it is described The longitudinal axis is greater than the first preset value；Or

If the human face posture is to face upward head and right side head, determine that the horizontal axis of the first area and the fourth region is less than The horizontal axis in first preset value, the second area and the third region is greater than first preset value, and described the The longitudinal axis of one region and the second area is greater than the first preset value, the third region and the fourth region it is described The longitudinal axis is less than the first preset value；Or

If the human face posture is to bow and left side head, determine that the horizontal axis of the first area and the fourth region is greater than The horizontal axis in first preset value, the second area and the third region is less than first preset value, and described the The longitudinal axis of one region and the second area is greater than the first preset value, the third region and the fourth region it is described The longitudinal axis is less than the first preset value；Or

If the human face posture is to face upward head and left side head, determine that the horizontal axis of the first area and the fourth region is less than The horizontal axis in first preset value, the second area and the third region is greater than first preset value, and described the The longitudinal axis of one region and the second area less than the first preset value, the third region and the fourth region it is described The longitudinal axis is greater than the first preset value.