WO2012002048A1 - 頭部検出方法、頭部検出装置、属性判定方法、属性判定装置、プログラム、記録媒体および属性判定システム - Google Patents
頭部検出方法、頭部検出装置、属性判定方法、属性判定装置、プログラム、記録媒体および属性判定システム Download PDFInfo
- Publication number
- WO2012002048A1 WO2012002048A1 PCT/JP2011/060981 JP2011060981W WO2012002048A1 WO 2012002048 A1 WO2012002048 A1 WO 2012002048A1 JP 2011060981 W JP2011060981 W JP 2011060981W WO 2012002048 A1 WO2012002048 A1 WO 2012002048A1
- Authority
- WO
- WIPO (PCT)
- Prior art keywords
- head
- image
- attribute determination
- head detection
- preliminary
- Prior art date
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/70—Determining position or orientation of objects or cameras
- G06T7/73—Determining position or orientation of objects or cameras using feature-based methods
- G06T7/74—Determining position or orientation of objects or cameras using feature-based methods involving reference images or patches
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/217—Validation; Performance evaluation; Active pattern learning techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10004—Still image; Photographic image
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30196—Human being; Person
- G06T2207/30201—Face
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30232—Surveillance
Definitions
- the present invention relates to a head detection method, a head detection device, an attribute determination method, an attribute determination device, a program, a recording medium, and an attribute determination system.
- a technique for detecting a person in an image using pattern recognition is performed, for example, by detecting part of a face (for example, eyes, nose, mouth, etc.), head, and skin color (for example, patents). Reference 1).
- scanning is performed by finely shifting an image patch in a predetermined area with respect to image data (original image) to be detected, and it is determined whether the image is a head image or the like. The determination is performed, for example, by referring to a learning image acquired in advance.
- an image patch that is captured at a predetermined position and size is a positive example head image, and a head that is out of position or size is negative (negative example).
- the size of the original image (detection target image) 70 can be changed in multiple stages, and scanning is performed for each size. In this case, heads of various sizes can be found by making the original image 70 smaller and searching while moving the image patch 71 in the same manner.
- the head may be missed.
- the patch is moved finely, it takes time for detection processing.
- the head may be missed unless the original image is gradually reduced. However, if the original image is gradually reduced, it still takes time for detection processing.
- the present invention provides a head detection method, a head detection device, an attribute determination method, an attribute determination device, a program, a recording medium, and an attribute determination system that detect a head in an image at high speed and accurately. Objective.
- the head detection method of the present invention comprises: A preliminary head detection model obtained as a positive example of an image including at least a part of the head in a predetermined image region defined in advance, and an image including no head as a negative example; A state in which the head is included in a state that matches a predetermined position and size in the predetermined image area as a positive example, and does not match at least one of the predetermined position and size And a deterministic head detection model obtained as a negative example of an image containing the head in An image acquisition step of acquiring a detection target image; A preliminary head detection step of cutting out the prescribed image region of the detection target image as an image patch and detecting a head image from the detection target image with reference to the preliminary head detection model; A deterministic head detection step of detecting a definite head image with reference to the deterministic head detection model from among the plurality of head images acquired by the preliminary head detection step. It is characterized by that.
- the head detecting device of the present invention is A preliminary head detection model obtained as a positive example of an image including at least a part of the head in a predetermined image region defined in advance, and an image including no head as a negative example; A state in which the head is included in a state that matches a predetermined position and size in the predetermined image area as a positive example, and does not match at least one of the predetermined position and size; A deterministic head detection model obtained as a negative image including the head in Image acquisition means for acquiring a detection target image; Preliminary head detection means for cutting out the prescribed image region of the detection target image as an image patch and detecting a head image from the detection target image with reference to the preliminary head detection model; Deterministic head detecting means for detecting a definite head image with reference to the deterministic head detection model from among the plurality of head images acquired by the preliminary head detecting means. It is characterized by that.
- the attribute determination method of the present invention includes: A head detection step of detecting the head by the head detection method of the present invention; An attribute determination step of determining an attribute from the image of the head.
- the attribute determination apparatus of the present invention is A head detecting means for detecting the head by the head detecting device of the present invention; And attribute determination means for determining an attribute from the image of the head.
- the program of the present invention causes a computer to execute at least one of the head detection method of the present invention and the attribute determination method of the present invention.
- the recording medium of the present invention records the program of the present invention.
- the attribute determination system of the present invention is Image acquisition means for acquiring a detection target image;
- a preliminary head detection model obtained as a positive example of an image including at least a part of the head in a predetermined image region defined in advance, and an image including no head as a negative example;
- a deterministic head detection model obtained as a negative image including the head in At least one of an attribute determination model and an attribute determination rule for determining an attribute from the image of the head;
- Preliminary head detection means for cutting out the prescribed image region of the detection target image as an image patch and detecting a head image from the detection target image with reference to the preliminary head detection model;
- Deterministic head detection means for detecting a definitive head image with reference to the deterministic head detection model from a plurality of head images acquired by the preliminary head detection means;
- a head detection method it is possible to provide a head detection method, a head detection device, an attribute determination method, an attribute determination device, a program, a recording medium, and an attribute determination system that detect a head in an image quickly and accurately. it can.
- FIG. 1A is a flowchart showing an example (Embodiment 1) of the head detection method of the present invention.
- FIG. 1B is a block diagram showing a configuration of an example (Embodiment 1) of the head detection device of the present invention.
- FIGS. 2A to 2F are diagrams for explaining an example of acquisition of a preliminary head detection model in the present invention.
- FIGS. 3A and 3B are diagrams illustrating an example of a preliminary head detection process in the head detection method of the first embodiment.
- FIGS. 3C and 3D are diagrams for explaining an example of a definitive head detection step in the head detection method of the first embodiment.
- FIG. 4A is a flowchart showing another example (embodiment 2) of the head detection method of the present invention.
- FIG. 1A is a flowchart showing an example (Embodiment 1) of the head detection method of the present invention.
- FIG. 1B is a block diagram showing a configuration of an example (Embodiment 1) of
- FIG. 4B is a block diagram showing a configuration of another example (Embodiment 2) of the head detecting apparatus of the present invention.
- FIG. 5A is a flowchart showing an example (third embodiment) of the attribute determination method of the present invention.
- FIG. 5B is a block diagram illustrating a configuration of an example (third embodiment) of the attribute determination apparatus of the present invention.
- FIG. 5C is a block diagram illustrating another example of the attribute determination apparatus according to the third embodiment.
- FIG. 6 is a block diagram showing a configuration of an example (embodiment 5) of the genus bottom determination system using the attribute determination device of the present invention.
- FIG. 7 is a diagram illustrating an example of a method for detecting a head in the related art.
- FIG. 1A shows a flowchart of a head detection method in the present embodiment.
- FIG. 1B shows a block diagram of the head detecting apparatus in the present embodiment.
- the head detecting apparatus of the present embodiment includes an image acquisition unit 111, a calculation unit 120, an output unit 131, and a data storage unit 140 as main components.
- the image acquisition unit 111 is electrically connected to the calculation unit 120.
- the computing means 120 is electrically connected to the output means 131 and the data storage means 140.
- the calculation means 120 includes a preliminary head detection means 121 and a definitive head detection means 122.
- the data storage means 140 stores a preliminary head detection model 141 and a definitive head detection model 142 acquired in advance.
- the preliminary head detection means 121 is connected to the preliminary head detection model 141.
- the definite head detection means 122 is connected to the deterministic head detection model 142.
- Examples of the image acquisition means 111 include a CCD (Charge Coupled Device) camera, a CMOS (Complementary Metal Oxide Semiconductor) camera, and an image scanner.
- An example of the calculation unit 120 is a central processing unit (CPU).
- Examples of the output means 131 include a monitor that outputs video (for example, various image display devices such as a liquid crystal display (LCD) and a cathode ray tube (CRT) display), a printer that outputs by printing, a speaker that outputs by sound, and the like. It is done.
- the output means 131 is an arbitrary component and may not be included in the head detecting device of the present invention, but is preferably included.
- Examples of the data storage means 140 include a random access memory (RAM), a read only memory (ROM), a hard disk (HD), an optical disk, a floppy (registered trademark) disk (FD), and the like.
- the data storage unit 140 may be, for example, a device built-in type or an external type such as an external storage device.
- the image acquisition unit, the calculation unit, the output unit, and the data storage unit are the same in the embodiments described later.
- the head detection method of the present embodiment is performed as follows using, for example, the head detection device of FIG. 1B.
- the head detection method learning is performed using machine learning (pattern recognition) technology. Specifically, first, an image including at least a part of the head in a predetermined image area defined in advance is used as a positive example from the learning image, and an image including no head is negative.
- the preliminary head detection model 141 is created by learning by machine learning (pattern recognition).
- the learning image (front-facing person image) 10A includes at least a part of the head in a prescribed image area (eg, 32 ⁇ 32 pixels) of a predetermined size. 10a (entire head part), 10b (upper right part of the head part), and 10c (left half part of the head part) are taken as positive examples.
- 10d person's shoulder
- 10e background
- the learning image (backward-facing person image) 10B is an image including at least a part of the head in the specified image region
- 10f head Overall
- 10 g left half of the head
- a learning image a human image that is captured in a small size facing the front
- 10C is an image that includes at least a part of the head in the prescribed image region.
- a certain example is 10h (entire head) and 10i (left half of the head).
- a learning image (a person image that is considerably large in the front direction) 10D is an image that includes at least a part of the head in the prescribed image region.
- a certain example is 10j (entire head) and 10k (lower left side of the head).
- the learning image (a human image that is reflected backwards in a very small size) 10E is an image that includes at least a part of the head in the prescribed image region.
- 10 m (entire head) and 10 n (upper half of the head) are positive examples.
- the learning image (a human image that is considerably rearward) 10F is an image that includes at least a part of the head in the prescribed image region.
- 10p (entire head) and 10q (lower right side of the head) are positive examples.
- an image that may be erroneously recognized as a head may be used as a negative example.
- the preliminary head detection model 141 is created by the machine learning as follows. First, a head region is accurately given to the head in the learning image so that there is no size shift and position shift (annotation).
- the head region is given by a human input using conventionally known input means such as a keyboard and a mouse.
- a case where the position of the head is shifted by 50% from the exact position in the head region is taken as a positive example, and for example, the size of the head is accurate in the head region.
- a preliminary range detection model 141 is created by designating an allowable range such as a positive example from the size up to ⁇ 30% and automatically creating a positive example by a program.
- an image including the head in a state that matches a predetermined position and size is taken as a positive example, and matches at least one of the predetermined position and size.
- the deterministic head detection model 142 is created by learning by machine learning (pattern recognition), taking an image including the head in a state where the head is not used as a negative example.
- machine learning pattern recognition
- an image in which the head is located approximately in the center of the defined image area and the outline (size) of the head is approximately the same size as the defined image area is defined as a positive example.
- images (10a and 10f) that meet the above definition are taken as positive examples.
- Images that do not meet the above rules (10b, 10c, 10g to 10k, 10m to 10n, and 10p to 10q) are taken as negative examples. In this way, it is only necessary to focus on negative examples of images that do not meet the above-mentioned rules from among the images that include the pre-detected heads. For example, images that do not include the heads, etc. It is not necessary to learn all of the above, and can be learned efficiently.
- the detection target image is acquired by the image acquisition unit 111 (step S11).
- the preliminary head detection means 121 cuts out the prescribed image region of the detection target image as an image patch, and refers to the preliminary head detection model 141 acquired in advance, and the head image is extracted from the detection target image. Is detected (step S21). Specifically, for example, as shown in FIG. 3A, the image patch 21 is moved in the horizontal direction from the upper left end of the detection target image 20 with reference to the preliminary head detection model 141 acquired in advance. The head image is searched by a so-called raster scan in which it is sequentially searched toward the lower row. In this example, the moving amount (width) of the image patch is set to 1 ⁇ 4 the size of the image patch. As a result, for example, head images 21a to 21g are detected as shown in FIG.
- the deterministic head detection unit 122 refers to the deterministic head detection model 142 from among the plurality of head images acquired in the preliminary head detection step S21, and determines the definite head image. Is detected (step S22). Specifically, in a slightly wide area including head images 21b, 21d, 21f, 21g and the like (images including the head of the left person in FIG. 3B) acquired in the preliminary head detection step S21. Is an input image. From this input image, with reference to the deterministic head detection model 142, a raster scan is performed and the input image is reduced, so that a definite head image 22a is obtained as shown in FIG. Is detected.
- an input image includes a slightly wider area including head images 21a, 21c and 21e acquired in the preliminary head detection step S21 (an image including the head of the right person in FIG. 3B). To do. From this input image, with reference to the deterministic head detection model 142, a raster scan is performed and the input image is reduced, so that a definite head image is obtained as shown in FIG. 22b is detected.
- the head detection result is output by the output means 131 (step S31).
- the output step S31 is an optional step and may not be included in the head detection method of the present invention, but is preferably included.
- the head detection method of the present embodiment first, an image including at least a part of the head is preliminarily detected from the detection target image. For this reason, even if the moving amount of the image patch is increased (for example, every 5 pixels) or the change rate of the image size is increased (for example, 0.8 times), the head is not missed. As a result, the head detection method of the present embodiment can detect candidates in the detection target image at high speed.
- a definite head image is detected from the head candidate images detected in advance. For this reason, definite head detection can also be performed at high speed. As a result, the head detection method of the present embodiment can detect the head in the detection target image with high speed and accuracy. About these effects, it is the same also in below-mentioned embodiment.
- FIG. 4A shows a flowchart of the head detection method in the present embodiment.
- FIG. 4B shows a block diagram of the head detecting apparatus in the present embodiment.
- the calculation means 120 replaces the preliminary head detecting means 121 with a preliminary head detecting means (first stage) 121-1 and a preliminary head.
- the head detection means (second stage) 121-2 is included, and the preliminary head detection model 141 in the data storage means 140 includes a first stage reference model 141-1 and a second stage reference model 141-2.
- the preliminary head detecting means (first stage) 121-1 is connected to the first stage reference model 141-1.
- the preliminary head detection means (second stage) 121-2 is connected to the second stage reference model 141-2.
- Other configurations are the same as those of the head detecting apparatus according to the first embodiment shown in FIG. 1B.
- the head detection method of the present embodiment is performed as follows using, for example, the head detection device of FIG. 4B.
- the preliminary head detection model 141 including the first stage reference model 141-1 and the second stage reference model 141-2 is created by learning by machine learning (pattern recognition).
- the learning shown in FIGS. 2 (a) to (f) is performed as in the preparation of the preliminary head detection model 141 in the first embodiment.
- images 10A to 10F include images 10a to 10c, 10f to 10k, 10m to 10n, and 10p to 10q that include at least a part of the head in the predetermined image area of the predetermined size.
- images 10d and 10e that do not include the head in the prescribed image region are taken as negative examples.
- the second-stage reference model 141-2 the case where the head is located at a position shifted by about half from the case where the head is located at the approximate center of the prescribed image region, and the outline (size) of the head. Is defined as a positive example from the case where the predetermined image area is approximately the same size to the image where the size is about half.
- images 10a to 10c, 10f to 10k, 10m to 10n, and 10p to 10q which are positive examples of the created first stage reference model 141-1
- images (10a, 10c, 10f, 10h) that meet the above definition 10i, 10m, and 10n) are positive examples
- images (10b, 10g, 10j, 10k, 10p, and 10q) that do not meet the above definition are negative examples.
- an image including at least a part of the head in the learning image is a positive example, but the present invention is not limited to this example,
- a positive example may be defined by specifying a predetermined allowable range for an accurate position and size in the head region.
- an image that satisfies both the position and size of the head is a positive example.
- the present invention is not limited to this example.
- an image that satisfies any one of the sizes may be a positive example. Therefore, in the first-stage reference model 141-1, for example, a case where the position of the head in the learning image is shifted by 50% from the exact position in the head region is used as a positive example.
- a positive example may be defined by specifying an allowable range such that the size of the part is from the exact size in the head region to ⁇ 30% as a positive example.
- the second-stage reference model 141-2 for example, up to 25% of the accurate position in the head region is taken as a positive example, and the size of the head is accurate in the head region.
- a positive example may be defined by specifying an allowable range such as a positive example from the size up to ⁇ 15%.
- the deterministic head detection model 142 is created in the same manner as in the first embodiment. Specifically, for example, an image in which the head is located approximately in the center of the defined image area and the outline (size) of the head is approximately the same size as the defined image area is defined as a positive example.
- images 10a, 10c, 10f, 10h, 10i, 10m, and 10n that are positive examples of the created second-stage reference model 141-2, images (10a and 10f) that meet the above definition are taken as positive examples. Images (10c, 10h, 10i, 10m, and 10n) that do not meet the above rules are taken as negative examples.
- the detection target image is acquired by the image acquisition unit 111 (step S11).
- the prescribed image region of the detection target image is cut out as an image patch by preliminary head detection means, and the first-stage reference model 141-1 and the second-stage reference model 141-2 acquired in advance are referred to.
- a head image is detected from the detection target image in multiple stages (two stages of the first stage and the second stage) (step S23).
- the first stage the first stage reference model 141-1 acquired in advance by the preliminary head detecting means (first stage) 121-1 is referred to in the first embodiment.
- a head image is searched from the detection target image 20 shown in FIG. 3A, and head images 21a to 21g, etc., as shown in FIG. Is detected.
- head images 21a, 21d, 21f and 21g are detected from the acquired images 21a to 21g with reference to the second stage reference model 141-2 acquired in advance. .
- the deterministic head detection unit 122 refers to the deterministic head detection model 142 from among the plurality of head images acquired in the preliminary head detection step S23. Then, a definite head image is detected (step S22). Specifically, in a slightly wide area including the head images 21d, 21f, 21g and the like (images including the head of the left person in FIG. 3B) acquired by the preliminary head detection step S23, Input image. From this input image, with reference to the deterministic head detection model 142, a raster scan is performed, and the input image is reduced, as shown in FIG. 22a is detected.
- a slightly wider area including the head image 21a and the like (image including the head of the right person in FIG. 3B) acquired from the preliminary head detection step S23 is set as an input image. From this input image, with reference to the deterministic head detection model 142, a raster scan is performed and the input image is reduced, so that a definite head image is obtained as shown in FIG. 22b is detected.
- the head detection result is output by the output means 131 in the same manner as in the first embodiment (step S31).
- the head is detected in two stages in the preliminary head detection process with reference to the reference model set in two stages. For this reason, the head in the detection target image can be detected more accurately at a higher speed.
- the reference model is set in two stages, and the preliminary head detection process is performed corresponding to this, but the present invention is not limited to this example, A reference model may be set in three or more stages, and a preliminary head detection process may be performed in accordance with this.
- FIG. 5A shows a flowchart of the attribute determination method in the present embodiment.
- FIG. 5B shows a block diagram of the attribute determination apparatus in the present embodiment.
- the attribute determination apparatus of this embodiment includes an image acquisition unit 111, a calculation unit 120, an output unit 131, and a data storage unit 140 as main components.
- the image acquisition unit 111 is electrically connected to the calculation unit 120.
- the computing means 120 is electrically connected to the output means 131 and the data storage means 140.
- the calculation unit 120 includes a preliminary head detection unit 121, a definitive head detection unit 122, and an attribute determination unit 124.
- the data storage unit 140 stores a preliminary head detection model 141, a deterministic head detection model 142, and an attribute determination model 144 acquired in advance.
- the preliminary head detection means 121 is connected to the preliminary head detection model 141.
- the definite head detection means 122 is connected to the deterministic head detection model 142.
- the attribute determination unit 124 is connected to the attribute determination model 144.
- a unit that combines the image acquisition unit 111, the preliminary head detection unit 121, and the definitive head detection unit 122 corresponds to the “head detection unit” in the present invention.
- the attributes are not particularly limited, and examples include sex, age, race, head orientation, hairstyle, hair length, presence / absence of a hat, and the like.
- the attribute determination method of the present embodiment is performed as follows using, for example, the attribute determination apparatus of FIG. 5B.
- the preliminary head detection model 141 and the deterministic head detection model 142 are created in the same manner as in the first embodiment.
- the head image 10a in the learning image 10A in FIG. 2A is a positive example in the deterministic head detection model 142.
- an attribute determination model 144 is created by machine learning (pattern recognition) using a large number of head images to which teacher data (or attribute values) is assigned.
- the teacher data (or attribute value) include sex, age, and the like. Specifically, for example, teacher data (or attribute values) such as “sex: male, age: 30 years old” is assigned to the head image 10a of the learning image 10A shown in FIG. Teacher data (or attribute values) such as sex and age are given by a person using a conventionally known input means such as a keyboard and a mouse. Then, the attribute determination model 144 is created using the head image 10a to which the teacher data (or attribute value) is assigned.
- the detection target image is acquired by the image acquisition unit 111 (step S11).
- the preliminary head detection unit 121 extracts the predetermined image area of the detection target image as an image patch, and refers to the preliminary head detection model 141 acquired in advance. Then, a head image is detected from the detection target image (step S21).
- the deterministic head detection unit 122 refers to the deterministic head detection model 142 from among the plurality of head images acquired in the preliminary head detection step S21. Then, a definite head image is detected (step S22). Specifically, for example, the definite head images 22a and 22b are detected as shown in FIGS. 3C and 3D in the same manner as in the first embodiment.
- a process combining the image acquisition process S11, the preliminary head detection process S21, and the definitive head detection process S22 corresponds to the “head detection process” in the present invention.
- the attribute is determined from the head image by the attribute determination unit 124 with reference to the attribute determination model 144 acquired in advance (step S24).
- the determination items include sex, age, head orientation, hairstyle, hair length, presence / absence of a hat, and the like.
- the determination item is gender, it can be determined based on, for example, the gender degree (for example, 0 to 1).
- the gender degree can be calculated based on, for example, a head image. Specifically, for example, if the gender degree is “0 to less than 0.5”, it is determined as “female”, and if the gender degree is “0.5 to 1”, it is determined as “male”.
- the gender is determined from the calculated gender degree value.
- the age and the like for example, a predetermined standard is set, and the age and the like are determined from a value calculated based on the head image.
- step S31 the attribute determination result is output by the output means 131 (step S31).
- the determination items are as described above.
- the output step S31 is an optional step and may not be included in the attribute determination method of the present invention, but is preferably included.
- the head is detected by the head detection method of the first embodiment, and the attribute is determined from the image of the head. Therefore, the attribute can be determined at high speed and accurately.
- the attribute is determined from the image of the head with reference to the attribute determination model. It is not limited to examples.
- the attribute determination may be performed with reference to an attribute determination rule, for example.
- Examples of the attribute determination rule include a rule such as “male if hair is short, female if hair is long”.
- the attribute determination may be performed with reference to both the attribute determination model and the attribute determination rule.
- the attribute determination rule 244 may be stored in the data storage unit 140, and the attribute determination unit 124 may be connected to the attribute determination rule 244.
- the preliminary head detection step in the attribute determination method of the present embodiment may be performed in multiple stages, for example, in the same manner as the preliminary head detection step in the head detection method of the second embodiment.
- the above-described multi-stage reference model is included in the preliminary head detection model in the attribute determination device. In this way, attributes can be determined more accurately at higher speed.
- the program of this embodiment is a program that can execute the above-described head detection method or the above-described attribute determination method on a computer.
- the program of this embodiment may be recorded on a recording medium, for example.
- the recording medium is not particularly limited, and examples thereof include a random access memory (RAM), a read only memory (ROM), a hard disk (HD), an optical disk, a floppy (registered trademark) disk (FD), and the like.
- FIG. 6 shows a configuration of an example of an attribute determination system using the attribute determination apparatus of the present invention.
- the attribute determination system includes image acquisition units 111a, 111b, and 111c, output units 131a, 131b, and 131c, communication interfaces 150a, 150b, and 150c, and a server 170.
- the image acquisition unit 111a and the output unit 131a are connected to the communication interface 150a.
- the image acquisition unit 111a, the output unit 131a, and the communication interface 150a are installed in the place X.
- the image acquisition unit 111b and the output unit 131b are connected to the communication interface 150b.
- the image acquisition unit 111b, the output unit 131b, and the communication interface 150b are installed at the place Y.
- the image acquisition unit 111c and the output unit 131c are connected to the communication interface 150c.
- the image acquisition unit 111c, the output unit 131c, and the communication interface 150c are installed in the place Z.
- Communication interfaces 150 a, 150 b, 150 c and server 170 are connected via a network 160.
- the server 170 side has preliminary head detection means, deterministic head detection means, and attribute determination means, and the server 170 has preliminary head detection models, deterministic head detection models, and attributes.
- the judgment model is stored. For example, the detection target image acquired at the place X using the image acquisition unit 111a is transmitted to the server 170, and the head is detected on the server 170 side, and the attribute is determined from the image of the head. The determined attribute is output by the output means 131a. Further, for example, the attribute determination rule may be stored in the server.
- the image acquisition means and the output means are installed at the site, and the server or the like is installed at another location, so that the head detection and attribute determination can be performed online. Therefore, for example, the installation of the apparatus does not take a place, and maintenance is easy. Further, for example, even when each installation place is separated, centralized management and remote operation at one place are possible.
- the attribute determination system of the present embodiment may be compatible with the multistage detection of the second embodiment described above. Moreover, the attribute determination system of this embodiment may be compatible with, for example, cloud computing.
- the present invention it is possible to provide a head detection method, a head detection device, an attribute determination method, an attribute determination device, a program, a recording medium, and an attribute determination system that detect a head in an image quickly and accurately.
- the invention can be applied to a wide range of uses.
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Artificial Intelligence (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Life Sciences & Earth Sciences (AREA)
- General Engineering & Computer Science (AREA)
- Image Analysis (AREA)
Abstract
Description
予め規定された規定画像領域において、頭部の少なくとも一部が含まれている画像を正例とし、かつ、頭部が含まれていない画像を負例として取得した予備的頭部検出モデルと、
前記規定画像領域において、予め規定された位置および大きさに合致する状態で頭部が含まれている画像を正例とし、かつ、前記予め規定された位置および大きさの少なくとも一方に合致しない状態で頭部が含まれている画像を負例として取得した確定的頭部検出モデルとを用い、
検出対象画像を取得する画像取得工程と、
前記検出対象画像の前記規定画像領域を画像パッチとして切り出し、前記予備的頭部検出モデルを参照して前記検出対象画像から頭部画像を検出する予備的頭部検出工程と、
前記予備的頭部検出工程により取得された複数の頭部画像の中から、前記確定的頭部検出モデルを参照して、確定的な頭部画像を検出する確定的頭部検出工程と
を含むことを特徴とする。
予め規定された規定画像領域において、頭部の少なくとも一部が含まれている画像を正例とし、かつ、頭部が含まれていない画像を負例として取得した予備的頭部検出モデルと、
前記規定画像領域において、予め規定された位置および大きさに合致する状態で頭部が含まれている画像を正例とし、かつ、前記予め規定された位置および大きさの少なくとも一方に合致しない状態で頭部が含まれている画像を負例として取得した確定的頭部検出モデルと、
検出対象画像を取得する画像取得手段と、
前記検出対象画像の前記規定画像領域を画像パッチとして切り出し、前記予備的頭部検出モデルを参照して前記検出対象画像から頭部画像を検出する予備的頭部検出手段と、
前記予備的頭部検出手段により取得された複数の頭部画像の中から、前記確定的頭部検出モデルを参照して、確定的な頭部画像を検出する確定的頭部検出手段と
を含むことを特徴とする。
前記本発明の頭部検出方法によって頭部を検出する頭部検出工程と、
前記頭部の画像から属性を判定する属性判定工程と
を含むことを特徴とする。
前記本発明の頭部検出装置によって頭部を検出する頭部検出手段と、
前記頭部の画像から属性を判定する属性判定手段と
を含むことを特徴とする。
検出対象画像を取得する画像取得手段と、
予め規定された規定画像領域において、頭部の少なくとも一部が含まれている画像を正例とし、かつ、頭部が含まれていない画像を負例として取得した予備的頭部検出モデルと、
前記規定画像領域において、予め規定された位置および大きさに合致する状態で頭部が含まれている画像を正例とし、かつ、前記予め規定された位置および大きさの少なくとも一方に合致しない状態で頭部が含まれている画像を負例として取得した確定的頭部検出モデルと、
前記頭部の画像から属性を判定するための属性判定モデルおよび属性判定ルールの少なくとも一方と、
前記検出対象画像の前記規定画像領域を画像パッチとして切り出し、前記予備的頭部検出モデルを参照して前記検出対象画像から頭部画像を検出する予備的頭部検出手段と、
前記予備的頭部検出手段により取得された複数の頭部画像の中から、前記確定的頭部検出モデルを参照して、確定的な頭部画像を検出する確定的頭部検出手段と、
前記頭部の画像から属性を判定する属性判定手段と、
属性判定結果を出力する出力手段と
を含み、
前記画像取得手段および前記出力手段が、システム外の通信回線網を介して、前記予備的頭部検出手段、前記予備的頭部検出モデル、前記確定的頭部検出手段、前記確定的頭部検出モデル、前記属性判定手段、ならびに、属性判定モデルおよび属性判定ルールの少なくとも一方と、接続されていることを特徴とする。
図1Aに、本実施形態における頭部検出方法のフローチャートを示す。また、図1Bに、本実施形態における頭部検出装置のブロック図を示す。図1Bに示すように、本実施形態の頭部検出装置は、画像取得手段111、演算手段120、出力手段131およびデータ記憶手段140を主要な構成要素として含む。画像取得手段111は、演算手段120に電気的に接続されている。演算手段120は、出力手段131とデータ記憶手段140とに電気的に接続されている。演算手段120は、予備的頭部検出手段121および確定的頭部検出手段122を含む。データ記憶手段140には、予め取得した予備的頭部検出モデル141および確定的頭部検出モデル142が格納されている。予備的頭部検出手段121は、予備的頭部検出モデル141に接続されている。確定的頭部検出手段122は、確定的頭部検出モデル142に接続されている。
図4Aに、本実施形態における頭部検出方法のフローチャートを示す。また、図4Bに、本実施形態における頭部検出装置のブロック図を示す。図4Bに示すように、本実施形態の頭部検出装置は、演算手段120が、予備的頭部検出手段121に代えて、予備的頭部検出手段(第1段階)121-1および予備的頭部検出手段(第2段階)121-2を含み、データ記憶手段140における予備的頭部検出モデル141が、第1段階参照用モデル141-1および第2段階参照用モデル141-2を含む。予備的頭部検出手段(第1段階)121-1は、第1段階参照用モデル141-1に接続されている。予備的頭部検出手段(第2段階)121-2は、第2段階参照用モデル141-2に接続されている。これら以外の構成は、図1Bに示す前記実施形態1の頭部検出装置と同様である。
図5Aに、本実施形態における属性判定方法のフローチャートを示す。また、図5Bに、本実施形態における属性判定装置のブロック図を示す。図5Bに示すように、本実施形態の属性判定装置は、画像取得手段111、演算手段120、出力手段131およびデータ記憶手段140を主要な構成要素として含む。画像取得手段111は、演算手段120に電気的に接続されている。演算手段120は、出力手段131とデータ記憶手段140とに電気的に接続されている。演算手段120は、予備的頭部検出手段121、確定的頭部検出手段122および属性判定手段124を含む。データ記憶手段140には、予め取得した予備的頭部検出モデル141、確定的頭部検出モデル142および属性判定モデル144が格納されている。予備的頭部検出手段121は、予備的頭部検出モデル141に接続されている。確定的頭部検出手段122は、確定的頭部検出モデル142に接続されている。属性判定手段124は、属性判定モデル144に接続されている。本実施形態の属性判定装置における、画像取得手段111、予備的頭部検出手段121および確定的頭部検出手段122を合わせた手段が、本発明における前記「頭部検出手段」に相当する。
本実施形態のプログラムは、前述の頭部検出方法または前述の属性判定方法を、コンピュータ上で実行可能なプログラムである。本実施形態のプログラムは、例えば、記録媒体に記録されてもよい。前記記録媒体としては、特に限定されず、例えば、ランダムアクセスメモリ(RAM)、読み出し専用メモリ(ROM)、ハードディスク(HD)、光ディスク、フロッピー(登録商標)ディスク(FD)等があげられる。
図6に、本発明の属性判定装置を用いた属性判定システムの一例の構成を示す。図6に示すとおり、この属性判定システムは、画像取得手段111a、111b、111cと、出力手段131a、131b、131cと、通信インターフェイス150a、150b、150cと、サーバ170とを備える。画像取得手段111aおよび出力手段131aは、通信インターフェイス150aに接続されている。画像取得手段111a、出力手段131aおよび通信インターフェイス150aは、場所Xに設置されている。画像取得手段111bおよび出力手段131bは、通信インターフェイス150bに接続されている。画像取得手段111b、出力手段131bおよび通信インターフェイス150bは、場所Yに設置されている。画像取得手段111cおよび出力手段131cは、通信インターフェイス150cに接続されている。画像取得手段111c、出力手段131cおよび通信インターフェイス150cは、場所Zに設置されている。そして、通信インターフェイス150a、150b、150cと、サーバ170とが、回線網160を介して接続されている。
10a、10b、10c、10d、10e、10f、10g、10h、10i、10j、10k、10m、10n、10p、10q 規定画像領域における画像
20、70 検出対象画像
21、71 画像パッチ
21a、21b、21c、21d、21e、21f、21g 頭部画像
22a、22b 確定的な頭部画像
111、111a、111b、111c 画像取得手段
120 演算手段
121 予備的頭部検出手段
121-1 予備的頭部検出手段(第1段階)
121-2 予備的頭部検出手段(第2段階)
122 確定的頭部検出手段
124 属性判定手段
131、131a、131b、131c 出力手段
140 データ記憶手段
141 予備的頭部検出モデル
141-1 第1段階参照用モデル
141-2 第2段階参照用モデル
142 確定的頭部検出モデル
144 属性判定モデル
150a、150b、150c 通信インターフェイス
160 回路網
170 サーバ
244 属性判定ルール
Claims (12)
- 予め規定された規定画像領域において、頭部の少なくとも一部が含まれている画像を正例とし、かつ、頭部が含まれていない画像を負例として取得した予備的頭部検出モデルと、
前記規定画像領域において、予め規定された位置および大きさに合致する状態で頭部が含まれている画像を正例とし、かつ、前記予め規定された位置および大きさの少なくとも一方に合致しない状態で頭部が含まれている画像を負例として取得した確定的頭部検出モデルとを用い、
検出対象画像を取得する画像取得工程と、
前記検出対象画像の前記規定画像領域を画像パッチとして切り出し、前記予備的頭部検出モデルを参照して前記検出対象画像から頭部画像を検出する予備的頭部検出工程と、
前記予備的頭部検出工程により取得された複数の頭部画像の中から、前記確定的頭部検出モデルを参照して、確定的な頭部画像を検出する確定的頭部検出工程と
を含むことを特徴とする頭部検出方法。 - 前記予備的頭部検出モデルが、前記確定的頭部検出モデルにおける予め規定された位置および大きさとの合致度合いが多段階に規定され、合致するものを正例とし、合致しないものを負例として取得した多段階の参照用モデルであり、
前記予備的頭部検出工程において、前記多段階の予備的頭部検出モデルの各段階に対応して頭部検出が多段階で実施され、
前の段階で取得された複数の頭部画像の中から、現段階の頭部画像を検出することを特徴とする、
請求項1記載の頭部検出方法。 - 予め規定された規定画像領域において、頭部の少なくとも一部が含まれている画像を正例とし、かつ、頭部が含まれていない画像を負例として取得した予備的頭部検出モデルと、
前記規定画像領域において、予め規定された位置および大きさに合致する状態で頭部が含まれている画像を正例とし、かつ、前記予め規定された位置および大きさの少なくとも一方に合致しない状態で頭部が含まれている画像を負例として取得した確定的頭部検出モデルと、
検出対象画像を取得する画像取得手段と、
前記検出対象画像の前記規定画像領域を画像パッチとして切り出し、前記予備的頭部検出モデルを参照して前記検出対象画像から頭部画像を検出する予備的頭部検出手段と、
前記予備的頭部検出手段により取得された複数の頭部画像の中から、前記確定的頭部検出モデルを参照して、確定的な頭部画像を検出する確定的頭部検出手段と
を含むことを特徴とする頭部検出装置。 - 前記予備的頭部検出モデルが、前記確定的頭部検出モデルにおける予め規定された位置および大きさとの合致度合いが多段階に規定され、合致するものを正例とし、合致しないものを負例として取得した多段階の参照用モデルであり、
前記予備的頭部検出手段が、前記多段階の予備的頭部検出モデルの各段階に対応して頭部検出を多段階で実施し、
前の段階で取得された複数の頭部画像の中から、現段階の頭部画像を検出することを特徴とする、
請求項3記載の頭部検出装置。 - 請求項1または2記載の頭部検出方法によって頭部を検出する頭部検出工程と、
前記頭部の画像から属性を判定する属性判定工程と
を含むことを特徴とする属性判定方法。 - 前記属性判定工程において、予め取得した属性判定モデルおよび属性判定ルールの少なくとも一方を参照して、前記頭部の画像から属性を判定することを特徴とする、請求項5記載の属性判定方法。
- 請求項3または4記載の頭部検出装置によって頭部を検出する頭部検出手段と、
前記頭部の画像から属性を判定する属性判定手段と
を含むことを特徴とする属性判定装置。 - 前記属性判定手段が、予め取得した属性判定モデルおよび属性判定ルールの少なくとも一方を参照して、前記頭部の画像から属性を判定することを特徴とする、請求項7記載の属性判定装置。
- 請求項1または2記載の頭部検出方法をコンピュータに実行させることを特徴とする、プログラム。
- 請求項5または6記載の属性判定方法をコンピュータに実行させることを特徴とする、プログラム。
- 請求項9または10記載のプログラムを記録していることを特徴とする記録媒体。
- 検出対象画像を取得する画像取得手段と、
予め規定された規定画像領域において、頭部の少なくとも一部が含まれている画像を正例とし、かつ、頭部が含まれていない画像を負例として取得した予備的頭部検出モデルと、
前記規定画像領域において、予め規定された位置および大きさに合致する状態で頭部が含まれている画像を正例とし、かつ、前記予め規定された位置および大きさの少なくとも一方に合致しない状態で頭部が含まれている画像を負例として取得した確定的頭部検出モデルと、
前記頭部の画像から属性を判定するための属性判定モデルおよび属性判定ルールの少なくとも一方と、
前記検出対象画像の前記規定画像領域を画像パッチとして切り出し、前記予備的頭部検出モデルを参照して前記検出対象画像から頭部画像を検出する予備的頭部検出手段と、
前記予備的頭部検出手段により取得された複数の頭部画像の中から、前記確定的頭部検出モデルを参照して、確定的な頭部画像を検出する確定的頭部検出手段と、
前記頭部の画像から属性を判定する属性判定手段と、
属性判定結果を出力する出力手段と
を含み、
前記画像取得手段および前記出力手段が、システム外の通信回線網を介して、前記予備的頭部検出手段、前記予備的頭部検出モデル、前記確定的頭部検出手段、前記確定的頭部検出モデル、前記属性判定手段、ならびに、属性判定モデルおよび属性判定ルールの少なくとも一方と、接続されていることを特徴とする属性判定システム。
Priority Applications (3)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201180032670.6A CN102971766B (zh) | 2010-06-30 | 2011-05-12 | 头部检测方法、头部检测装置、属性判定方法、属性判定装置和属性判定系统 |
US13/807,677 US8917915B2 (en) | 2010-06-30 | 2011-05-12 | Head detecting method, head detecting apparatus, attribute determining method, attribute determining apparatus, program, recording medium, and attribute determining system |
JP2012522506A JP5451883B2 (ja) | 2010-06-30 | 2011-05-12 | 頭部検出方法、頭部検出装置、属性判定方法、属性判定装置、プログラム、記録媒体および属性判定システム |
Applications Claiming Priority (2)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
JP2010149934 | 2010-06-30 | ||
JP2010-149934 | 2010-06-30 |
Publications (1)
Publication Number | Publication Date |
---|---|
WO2012002048A1 true WO2012002048A1 (ja) | 2012-01-05 |
Family
ID=45401783
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
PCT/JP2011/060981 WO2012002048A1 (ja) | 2010-06-30 | 2011-05-12 | 頭部検出方法、頭部検出装置、属性判定方法、属性判定装置、プログラム、記録媒体および属性判定システム |
Country Status (4)
Country | Link |
---|---|
US (1) | US8917915B2 (ja) |
JP (1) | JP5451883B2 (ja) |
CN (1) | CN102971766B (ja) |
WO (1) | WO2012002048A1 (ja) |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2020128326A1 (fr) | 2018-12-20 | 2020-06-25 | Bostik Sa | Composition à base de résine époxy et de polyuréthane |
KR20200128565A (ko) * | 2018-07-23 | 2020-11-13 | 텐센트 테크놀로지(센젠) 컴퍼니 리미티드 | 이미지 처리 방법 및 장치, 단말 및 컴퓨터 판독 가능 저장 매체 |
Families Citing this family (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2017170440A1 (ja) * | 2016-03-28 | 2017-10-05 | Necソリューションイノベータ株式会社 | 計測装置、計測方法、及びコンピュータ読み取り可能な記録媒体 |
WO2019056102A1 (en) * | 2017-09-19 | 2019-03-28 | Intuitive Robotics, Inc. | SYSTEMS AND METHODS FOR DETECTION AND RECOGNITION OF WASTE ARTICLES |
Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2006073081A1 (ja) * | 2005-01-05 | 2006-07-13 | Nec Corporation | 識別用データ学習システム、学習装置、識別装置及び学習方法 |
JP2009059047A (ja) * | 2007-08-30 | 2009-03-19 | Victor Co Of Japan Ltd | 対象物検出装置、対象物検出方法、および対象物検出プログラム |
WO2010032298A1 (ja) * | 2008-09-17 | 2010-03-25 | 富士通株式会社 | 画像処理装置、画像処理方法および画像処理プログラム |
Family Cites Families (11)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP3810943B2 (ja) | 1999-05-06 | 2006-08-16 | 株式会社東芝 | 画像処理装置、画像処理方法および画像処理プログラムを記録した記録媒体 |
US6940545B1 (en) * | 2000-02-28 | 2005-09-06 | Eastman Kodak Company | Face detecting camera and method |
US7657083B2 (en) * | 2000-03-08 | 2010-02-02 | Cyberextruder.Com, Inc. | System, method, and apparatus for generating a three-dimensional representation from one or more two-dimensional images |
EP1639522B1 (en) * | 2003-06-30 | 2007-08-15 | HONDA MOTOR CO., Ltd. | System and method for face recognition |
KR100624481B1 (ko) * | 2004-11-17 | 2006-09-18 | 삼성전자주식회사 | 형판 기반 얼굴 검출 방법 |
JP4619762B2 (ja) * | 2004-12-10 | 2011-01-26 | 富士フイルム株式会社 | 画像処理方法および装置並びにプログラム |
JP4386447B2 (ja) * | 2005-09-26 | 2009-12-16 | 富士フイルム株式会社 | 画像分割装置および方法並びにプログラム |
US7889892B2 (en) * | 2005-10-13 | 2011-02-15 | Fujifilm Corporation | Face detecting method, and system and program for the methods |
CN100464332C (zh) * | 2007-03-20 | 2009-02-25 | 北京中星微电子有限公司 | 一种图片查询方法及系统 |
JP4881199B2 (ja) * | 2007-03-23 | 2012-02-22 | 富士フイルム株式会社 | 画像評価装置および方法並びにプログラム |
US8818034B2 (en) * | 2009-11-30 | 2014-08-26 | Hewlett-Packard Development Company, L.P. | Face recognition apparatus and methods |
-
2011
- 2011-05-12 CN CN201180032670.6A patent/CN102971766B/zh not_active Expired - Fee Related
- 2011-05-12 JP JP2012522506A patent/JP5451883B2/ja active Active
- 2011-05-12 US US13/807,677 patent/US8917915B2/en active Active
- 2011-05-12 WO PCT/JP2011/060981 patent/WO2012002048A1/ja active Application Filing
Patent Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2006073081A1 (ja) * | 2005-01-05 | 2006-07-13 | Nec Corporation | 識別用データ学習システム、学習装置、識別装置及び学習方法 |
JP2009059047A (ja) * | 2007-08-30 | 2009-03-19 | Victor Co Of Japan Ltd | 対象物検出装置、対象物検出方法、および対象物検出プログラム |
WO2010032298A1 (ja) * | 2008-09-17 | 2010-03-25 | 富士通株式会社 | 画像処理装置、画像処理方法および画像処理プログラム |
Cited By (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
KR20200128565A (ko) * | 2018-07-23 | 2020-11-13 | 텐센트 테크놀로지(센젠) 컴퍼니 리미티드 | 이미지 처리 방법 및 장치, 단말 및 컴퓨터 판독 가능 저장 매체 |
US11631275B2 (en) | 2018-07-23 | 2023-04-18 | Tencent Technology (Shenzhen) Company Limited | Image processing method and apparatus, terminal, and computer-readable storage medium |
KR102635373B1 (ko) * | 2018-07-23 | 2024-02-07 | 텐센트 테크놀로지(센젠) 컴퍼니 리미티드 | 이미지 처리 방법 및 장치, 단말 및 컴퓨터 판독 가능 저장 매체 |
WO2020128326A1 (fr) | 2018-12-20 | 2020-06-25 | Bostik Sa | Composition à base de résine époxy et de polyuréthane |
FR3090672A1 (fr) | 2018-12-20 | 2020-06-26 | Bostik Sa | Composition à base de résine époxy et de polyuréthane |
Also Published As
Publication number | Publication date |
---|---|
US20130114889A1 (en) | 2013-05-09 |
CN102971766A (zh) | 2013-03-13 |
US8917915B2 (en) | 2014-12-23 |
JP5451883B2 (ja) | 2014-03-26 |
CN102971766B (zh) | 2016-06-29 |
JPWO2012002048A1 (ja) | 2013-08-22 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
US11200404B2 (en) | Feature point positioning method, storage medium, and computer device | |
JP5569990B2 (ja) | 属性判定方法、属性判定装置、プログラム、記録媒体および属性判定システム | |
US20220148337A1 (en) | Living body detection method and apparatus, electronic device, and storage medium | |
JP5213105B2 (ja) | 映像ネットワークシステム及び映像データ管理方法 | |
WO2019047789A1 (zh) | 关于增强现实场景的处理方法、终端设备、系统及计算机存储介质 | |
US10699473B2 (en) | System and method for generating a virtual viewpoint apparatus | |
US20180114363A1 (en) | Augmented scanning of 3d models | |
WO2021012370A1 (zh) | 瞳孔半径的检测方法、装置、计算机设备和存储介质 | |
EP3709266A1 (en) | Human-tracking methods, apparatuses, systems, and storage media | |
WO2018133825A1 (zh) | 视频通话中视频图像的处理方法、终端设备、服务器及存储介质 | |
KR20140139730A (ko) | 깊이 차이를 이용한 얼굴 구성요소의 자동 분류 방법 | |
CN104811660A (zh) | 控制装置及控制方法 | |
US20160323505A1 (en) | Photographing processing method, device and computer storage medium | |
JP6157165B2 (ja) | 視線検出装置及び撮像装置 | |
JP7064257B2 (ja) | 画像深度確定方法及び生き物認識方法、回路、装置、記憶媒体 | |
CN112116582A (zh) | 一种库存或陈列场景下的条烟检测识别方法 | |
CN113221767B (zh) | 训练活体人脸识别模型、识别活体人脸的方法及相关装置 | |
JP5451883B2 (ja) | 頭部検出方法、頭部検出装置、属性判定方法、属性判定装置、プログラム、記録媒体および属性判定システム | |
JP2018026115A (ja) | 火炎検出方法、火炎検出装置及び電子機器 | |
CN109002776B (zh) | 人脸识别方法、系统、计算机设备和计算机可读存储介质 | |
JP6290020B2 (ja) | 画像処理装置、画像処理方法、およびプログラム | |
CN112184837A (zh) | 一种图像检测方法、装置、电子设备及存储介质 | |
JP2016099643A (ja) | 画像処理装置、画像処理方法および画像処理プログラム | |
CN103473807A (zh) | 一种3d模型变换系统及方法 | |
KR101360063B1 (ko) | 제스처 인식 방법 및 시스템 |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
WWE | Wipo information: entry into national phase |
Ref document number: 201180032670.6 Country of ref document: CN |
|
121 | Ep: the epo has been informed by wipo that ep was designated in this application |
Ref document number: 11800517 Country of ref document: EP Kind code of ref document: A1 |
|
WWE | Wipo information: entry into national phase |
Ref document number: 2012522506 Country of ref document: JP |
|
WWE | Wipo information: entry into national phase |
Ref document number: 13807677 Country of ref document: US |
|
NENP | Non-entry into the national phase |
Ref country code: DE |
|
122 | Ep: pct application non-entry in european phase |
Ref document number: 11800517 Country of ref document: EP Kind code of ref document: A1 |