CN110718068B - A method for estimating the installation angle of a road surveillance camera - Google Patents

Abstract

The invention discloses a method for estimating the installation angle of a road monitoring camera, which belongs to the field of road monitoring. Including: detecting the horizontal vanishing point based on parallel lines on the image obtained by the camera; using the relationship between the horizontal vanishing point in the image and the installation angle of the camera to obtain the installation angle of the road surveillance camera. The invention combines the prior knowledge of the external conditions of the road monitoring scene and the projection transformation principle in the imaging principle, utilizes the relationship between the vanishing point and the installation angle of the camera, and is supplemented by the fact that there are only two degrees of freedom in the installation process of the security camera. Features, the two installation angles of interest are calculated through mathematical formulas. Therefore, after the surveillance camera is installed, the installation angle of the camera can be automatically calculated without adding any operation process. Automatically identify the installation angle at a fixed time interval to adapt to the influence of natural environment factors on the installation angle of the camera, and achieve the purpose of automatic labeling.

Description

A method for estimating the installation angle of a road surveillance camera

technical field

The invention belongs to the field of road monitoring, and more particularly, relates to a method for estimating the installation angle of a road monitoring camera.

Background technique

In recent years, with the rapid economic development, the number of motor vehicles in urban and rural areas in my country has increased rapidly, and criminal and public security cases related to vehicles have also increased year by year. Under this circumstance, it has become an important task for public security traffic management departments at all levels to effectively use advanced scientific and technological means to improve urban and rural road traffic management, combat and prevent vehicle-related cases, deter criminals, and improve social security management. Many functions of road surveillance cameras need to calibrate the external parameters of the camera, such as the height of the camera installation, the pitch angle, yaw angle and roll angle of the camera installation, etc. In the process of camera deployment, it is difficult to measure the installation angle. Moreover, in road monitoring, the installation angle of the camera may also be changed by the influence of the external environment, such as wind blowing, pole shaking, etc., resulting in inaccuracies behind the angle measured in the previous stage.

At present, the measurement of the camera installation angle generally has the following measurement methods: 1. Measure by angle measuring tools such as level; 2. Measure by installing gyroscope and other equipment in the camera; 3. Manually find the reference object in the camera, Then the camera installation angle is calculated through the transformation relationship.

However, the first method is to measure the angle through a measuring tool, which is very difficult to implement and cannot be automatically updated after one measurement. In the second method, there are very few security surveillance cameras such as gyroscopes, so it is difficult to spread them out in a large area, and adding one more device means increasing costs and reducing system stability. The third method requires a clear reference object in the camera's field of view, which is not necessarily available in most cameras. Moreover, it can only be calibrated once during installation. Security surveillance cameras work in outdoor scenes all year round. Wind blowing, high temperature deformation, etc. may cause the installation angle of the camera to change, and this method cannot automatically identify the installation angle to adapt to these application scenarios. .

SUMMARY OF THE INVENTION

Aiming at the shortcomings of the prior art method of measuring the angle of high cost and inability to adapt, the present invention provides a method for estimating the installation angle of a road monitoring camera, the purpose of which is to automatically calibrate the installation angle of the camera without human intervention. The distance measurement and speed measurement of the camera provide automatically extracted parameters to automatically calibrate the installation angle of the camera in real time at the lowest cost.

In order to achieve the above object, according to the first aspect of the present invention, a method for estimating the installation angle of a road monitoring camera is provided, and the method includes the following steps:

S1. For the image acquired by the camera, detect the horizontal vanishing point based on parallel lines;

S2. Use the relationship between the horizontal vanishing point in the image and the installation angle of the camera to obtain the installation angle of the road surveillance camera.

Specifically, step S1 includes the following sub-steps:

S11. Detect straight line segments in the enhanced image;

S12. remove the horizontal line segment from the detected line segment set;

S13. Use the remaining straight line segments to find the intersection points two by two, and obtain the intersection point area of the straight line segments;

S14. Obtain the center point of the intersection area, and use the center point as the horizontal vanishing point of the image;

S15. Using the width, height and image center information of the image, normalize the vanishing point to obtain the normalized vanishing point.

Specifically, step S1 includes the following sub-steps:

S111. Obtain the Y channel from the input image, and perform lane line enhancement based on the Y channel;

S112. Use the LSD algorithm to detect straight line segments in the enhanced image.

Specifically, the lane line enhancement based on the Y channel is as follows:

E(x,y)=2P(x,y)-P(x-N,y)-P(x+N,y)

Among them, E(x,y) represents the pixel whose abscissa is x and ordinate is y after lane line enhancement, and P(x,y) represents the pixel whose abscissa is x and ordinate is y in Y channel before lane line enhancement , the width of the lane line in the image is 2*N.

Specifically, in step S12, whether the straight line segment is a horizontal line is judged by the included angle between the straight line segment and the horizontal: if the absolute value of the included angle between the straight line segment and the horizontal direction of the image is less than α, the straight line segment is considered to be a horizontal line segment, The value range of the threshold α is [10°, 30°].

Specifically, in step S14, the MeanShift algorithm is used to obtain the center point of the intersection area.

Specifically, the use of the relationship between the horizontal vanishing point in the image and the installation angle of the camera to obtain the installation angle of the road surveillance camera is as follows:

表示俯仰角，Ψ表示滚动角，vpx和vpy分别表示图像中的水平消失点在归一化图像中的位置。where Φ represents the yaw angle,

represents the pitch angle, Ψ represents the roll angle, and vpx and vpy represent the position of the horizontal vanishing point in the image in the normalized image, respectively.

Specifically, for the image stream obtained by the camera, steps S1 to S2 are repeated at regular intervals to realize automatic identification of the installation angle of the road monitoring camera.

In order to achieve the above object, according to a second aspect of the present invention, a computer-readable storage medium is provided, and a computer program is stored on the computer-readable storage medium, and when the computer program is executed by a processor, the first aspect is implemented The method for estimating the installation angle of the road monitoring camera.

In general, through the above technical solutions conceived by the present invention, the following beneficial effects can be achieved:

The invention combines the prior knowledge of the external conditions of the road monitoring scene and the projection transformation principle in the imaging principle, uses the relationship between the vanishing point and the installation angle of the camera, and supplements it with two degrees of freedom in the installation process of the security camera. (that is, the roll angle of the camera is 0), and the two installation angles of interest are calculated by mathematical formulas. Therefore, after the monitoring camera is installed, the installation angle of the camera can be automatically calculated without adding any operation process, and the installed security monitoring camera is also applicable. Since the method automatically calculates the installation angle, it can automatically identify the installation angle at a fixed time interval to adapt to the influence of natural environment factors on the installation angle of the camera, and achieve the purpose of automatic labeling.

Description of drawings

1 is a flowchart of a method for estimating an installation angle of a road surveillance camera according to an embodiment of the present invention;

2 is a schematic diagram of a vanishing point in an image provided by an embodiment of the present invention;

FIG. 3 is a flowchart of vanishing point detection provided by an embodiment of the present invention.

Detailed ways

In order to make the objectives, technical solutions and advantages of the present invention clearer, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention, but not to limit the present invention. In addition, the technical features involved in the various embodiments of the present invention described below can be combined with each other as long as there is no conflict with each other.

The invention combines the prior knowledge of the external conditions of the road monitoring scene and the projection transformation principle in the imaging principle, and proposes an algorithm for automatic identification of the installation angle of the road monitoring camera, which can automatically calibrate the camera without human intervention. The installation angle provides automatically extracted parameters for the distance measurement and speed measurement of the 3D camera.

As shown in FIG. 1 , the present invention provides a method for estimating the installation angle of a road surveillance camera, which includes the following steps:

Step S1. For the image acquired by the camera, detect the horizontal vanishing point based on parallel lines.

This method is applied to road security surveillance cameras, mainly cameras without roll angles, such as bolts and ball cameras. The vanishing point in the image refers to the intersection of the parallel lines of the physical world in the image as shown in Figure 2. Therefore, the detection of the vanishing point in the image requires the existence of parallel lines in the image, and there are many parallel lines in road monitoring. Such as road traffic signs, road edges, fences, etc., can be used as raw materials for detecting vanishing points.

As shown in Figure 3, step S1 includes the following sub-steps:

Step S11. Detect straight line segments in the enhanced image.

S111. Obtain the Y channel from the input image and use lane line enhancement.

Using the inherent characteristics of road monitoring, the road signs are enhanced in the road monitoring images to facilitate the extraction of straight line segments from them.

The gray value of all lane lines in the Y channel is higher than the gray value of the background pixels of the road surface, and the lane lines are all strip-shaped, and the width of the lane lines is relatively fixed. Therefore, based on this characteristic, the present invention adopts the following method to enhance the lane line:

Assuming that the width of the lane line in the image is 2*N, use P(x,y) to represent the pixel whose abscissa is x and the ordinate is y in the Y channel before lane line enhancement, and E(x,y) represents the lane line enhancement After the pixel whose abscissa is x and its ordinate is y, then E(x,y)=2P(x,y)-P(x-N,y)-P(x+N,y). After being enhanced by this method, the part of the lane line will be highlighted in the image.

S112. Use an LSD (Line_Segment_Detector, straight line segment detection) algorithm to detect straight line segments in the enhanced image.

There are a lot of parallel lines available in the road monitoring scene. Using computer vision knowledge, the parallel lines can be automatically extracted.

The present invention preferably detects the straight line segment in the enhanced image by the LSD algorithm, and the straight line detected by the LSD algorithm is more accurate, and is more suitable for lane lines with alternate virtual and real.

Step S12. Remove the horizontal line segment from the detected line segment set.

Because of the need to find the vanishing point, it is necessary to eliminate the influence of straight line segments in the horizontal or near-horizontal direction. Whether the straight line segment is a straight line in the horizontal direction is judged by the included angle between the straight line segment and the horizontal. The horizontal line segment is removed from the detected straight line segment, so that the intersection point (ie, the vanishing point) of the parallel lines of the physical world in the image is obtained by using the parallel line characteristic of road traffic signs.

If the absolute value of the angle between the straight line segment and the horizontal direction of the image is less than 15°, it is considered as a horizontal line segment; if the absolute value of the angle between the straight line segment and the vertical direction is less than 15°, it is determined as a vertical line segment.

Step S13. Use the remaining straight line segments to find the intersection points two by two, and obtain the intersection point area of the straight line segments.

Step S14. Obtain the center point of the intersection area, and use the center point as the horizontal vanishing point of the image.

The vanishing point in the image can be estimated using the parallel lines of the physical world.

In the present invention, the MeanShift algorithm is preferred to obtain the center point of the intersection area, and the obtained center point has higher accuracy, is less affected by initialization, and is less disturbed by noise.

Step S15. Use the width, height and image center information of the image to perform a normalization operation on the vanishing point to obtain a normalized vanishing point.

Step S2. Obtain the installation angle of the road surveillance camera by using the relationship between the horizontal vanishing point in the image and the installation angle of the camera.

Using the relationship between the vanishing point and the camera angle in the imaging principle, and assuming that the camera roll angle is 0, the equation with infinite solutions can be transformed into an equation with feasible solutions, so as to obtain the angle of the camera installation.

Using the imaging principle, the relationship between the horizontal vanishing point in the image and the installation angle of the camera is shown in the following formula:

表示俯仰角，Ψ表示滚动角，一般道路监控的相机支架只有两个角度的自由度，即俯仰角、偏航角两个角度，因此只要将相机装正，就能保证相机的滚动角为0°，将该已知条件代入上面的公式，可以得到如下的公式：where Φ represents the yaw angle,

Indicates the pitch angle, and Ψ indicates the roll angle. Generally, the camera bracket for road monitoring has only two degrees of freedom, namely the pitch angle and the yaw angle. Therefore, as long as the camera is installed upright, the roll angle of the camera can be guaranteed to be 0. °, substitute this known condition into the above formula, you can get the following formula:

Among them, vpx and vpy respectively represent the position of the horizontal vanishing point in the image in the normalized image. There are only two unknowns in this equation, and two equations can be formed, so the equation has a solution, and the equation can be obtained by solving the equation:

It can be known from the above equation that the three-dimensional installation angle of the camera can be obtained only by locating the normalized position of the horizontal vanishing point in the image. The three-dimensional installation angle of the camera can be expressed as:

Those skilled in the art can easily understand that the above are only preferred embodiments of the present invention, and are not intended to limit the present invention. Any modifications, equivalent replacements and improvements made within the spirit and principles of the present invention, etc., All should be included within the protection scope of the present invention.

Claims (8)

1. a road monitoring camera installation angle estimation method, is characterized in that, the method comprises the following steps: S1. For the image acquired by the camera, detect the horizontal vanishing point based on parallel lines; S2. Use the relationship between the horizontal vanishing point in the image and the installation angle of the camera to obtain the installation angle of the road surveillance camera, as follows:

where Φ represents the yaw angle, Θ represents the pitch angle, Ψ represents the roll angle, and vpx and vpy represent the position of the horizontal vanishing point in the image in the normalized image, respectively. 2. The method of claim 1, wherein step S1 comprises the following substeps: S11. Detect straight line segments in the enhanced image; S12. remove the horizontal line segment from the detected line segment set; S13. Use the remaining straight line segments to find the intersection points two by two, and obtain the intersection point area of the straight line segments; S14. Obtain the center point of the intersection area, and use the center point as the horizontal vanishing point of the image; S15. Use the width, height and image center information of the image to perform a normalization operation on the vanishing point to obtain a normalized vanishing point. 3. The method of claim 2, wherein step S1 comprises the following substeps: S111. Obtain the Y channel from the input image, and perform lane line enhancement based on the Y channel; S112. Use the LSD algorithm to detect straight line segments in the enhanced image. 4. The method of claim 3, wherein the lane line enhancement based on the Y channel is as follows: E(x,y)=2P(x,y)-P(x-N,y)-P(x+N,y) Among them, E(x,y) represents the pixel whose abscissa is x and ordinate is y after lane line enhancement, and P(x,y) represents the pixel whose abscissa is x and ordinate is y in Y channel before lane line enhancement , the width of the lane line in the image is 2*N. 5. The method according to claim 2, wherein in step S12, it is judged whether the straight line segment is a straight line in the horizontal direction by the included angle between the straight line segment and the horizontal direction: if the absolute value of the included angle between the straight line segment and the horizontal direction of the image If it is less than α, the straight line segment is considered to be a horizontal line segment, and the value range of the threshold α is [10°, 30°]. 6 . The method of claim 2 , wherein, in step S14 , the MeanShift algorithm is used to obtain the center point of the intersection area. 7 . 7 . The method of claim 1 , wherein steps S1 to S2 are repeated at regular intervals for the image stream obtained by the camera to realize automatic identification of the installation angle of the road surveillance camera. 8 . 8. A computer-readable storage medium, wherein a computer program is stored on the computer-readable storage medium, and when the computer program is executed by a processor, the road according to any one of claims 1 to 7 is implemented Surveillance camera installation angle estimation method.

Images