CN113139504B

CN113139504B - Identity recognition method, device, equipment and storage medium

Info

Publication number: CN113139504B
Application number: CN202110511744.XA
Authority: CN
Inventors: 韩煦深; 赵雄心
Original assignee: Alipay Hangzhou Information Technology Co Ltd
Current assignee: Alipay Hangzhou Information Technology Co Ltd
Priority date: 2021-05-11
Filing date: 2021-05-11
Publication date: 2023-02-17
Anticipated expiration: 2041-05-11
Also published as: CN113139504A

Abstract

The embodiment of the specification provides an identity recognition method and an identity recognition device, wherein the identity recognition method comprises the following steps: the method comprises the steps of detecting key points of an object to be recognized contained in an obtained image to be processed, determining a feature detection area corresponding to the object to be recognized according to a key point detection result, extracting image features of the feature detection area, carrying out pedestrian re-recognition on the object to be recognized by utilizing the image features, and generating an identity recognition result corresponding to the object to be recognized.

Description

Identity recognition method, device, equipment and storage medium

Technical Field

The embodiment of the specification relates to the technical field of computers, in particular to an identity recognition method. One or more embodiments of the present specification also relate to an identification apparatus, a computing device, and a computer-readable storage medium.

Background

With the development of video technology, cameras are deployed in more and more scenes and can be used for collecting scene pictures, such as hospitals, markets, railway stations and the like, and the collected scene pictures can provide a powerful data basis for security detection and analysis in the fields of security, commerce and the like.

The pedestrian re-identification technology is an important research field in image retrieval, is suitable for carrying out key point feature analysis and identification on pedestrians in different scenes, and gradually increases the analysis result of each feature to the global human body feature matching result so as to finish the accurate identification of special people.

However, the existing pedestrian re-identification method mainly directly extracts image features in a rectangular frame where a user is located, and judges the identity of a person by using the distance between the image features, and if the person is blocked among the persons or one main person and a plurality of other persons are included in one rectangular frame due to the angle problem of a camera, the accuracy of an identification result is low, so that an effective method is urgently needed to solve the problem.

Disclosure of Invention

In view of this, the embodiments of the present specification provide an identity recognition method. One or more embodiments of the present disclosure also relate to an identification apparatus, a computing device, and a computer-readable storage medium to solve the technical problems in the prior art.

According to a first aspect of embodiments of the present specification, there is provided an identity recognition method, including:

detecting key points of an object to be identified contained in the acquired image to be processed;

determining a feature detection area corresponding to the object to be identified according to a key point detection result;

and extracting the image characteristics of the characteristic detection area, and carrying out pedestrian re-identification on the object to be identified by utilizing the image characteristics to generate an identity identification result corresponding to the object to be identified.

Optionally, the performing of the key point detection on the object to be identified included in the acquired image to be processed includes:

inputting the acquired image to be processed into a key point detection model, wherein the key point detection model is used for detecting key points of an object to be identified contained in the image to be processed;

the key point detection model is trained in the following way:

acquiring a sample image and a key point marking result of an object to be identified in the sample image;

and training the key point detection model by taking the sample image as a training sample and taking the key point labeling result as a label to obtain the key point detection model.

Optionally, the determining, according to the key point detection result, the feature detection area corresponding to the object to be identified includes:

determining whether the key point detection result contains a first target key point and a second target key point;

and if so, determining the feature detection area of the object to be identified based on the connecting line between the first target key point and the second target key point and the key point detection result.

if the fact that the key point detection result contains the first target key point is determined, establishing a two-dimensional image coordinate system by taking any vertex of the image to be processed as a coordinate origin;

determining a first coordinate of the first target key point in the two-dimensional image coordinate system;

calculating a reference coordinate of the first target key point in a three-dimensional reference coordinate system according to a coordinate conversion relation between a two-dimensional image coordinate system and the three-dimensional reference coordinate system and the first coordinate;

calculating a second coordinate of any point on a straight line which passes through the reference coordinate and is parallel to a vertical axis of the three-dimensional reference coordinate system in the image coordinate system according to the reference coordinate and the coordinate conversion relation;

and determining a characteristic detection area of the object to be identified based on a connecting line between the first target key point and the coordinate point of the second coordinate and the key point detection result.

if the first target key point is determined not to be contained in the key point detection result, establishing a two-dimensional image coordinate system by taking any vertex of the image to be processed as a coordinate origin;

determining a first coordinate of a second target key point in the two-dimensional image coordinate system;

calculating the reference coordinate of the second target key point in the three-dimensional reference coordinate system according to the coordinate conversion relation between the two-dimensional image coordinate system and the three-dimensional reference coordinate system, the first coordinate and the preset value corresponding to the second target key point in the vertical axis direction of the three-dimensional reference coordinate system;

and determining the characteristic detection area of the object to be identified based on the connecting line between the second target key point and the coordinate point of the second coordinate and the key point detection result.

if so, determining a connecting line between the first target key point and the second target key point, and determining a rotation angle and a rotation direction of an initial feature detection area of the object to be identified in the image to be processed according to an angle and a position relation between the connecting line and the edge of the image to be processed;

and rotating the initial feature detection area according to the rotation angle and the rotation direction to generate a feature detection area corresponding to the object to be identified.

determining a connecting line of the first target key point and a coordinate point of a second coordinate, and determining a rotation angle and a rotation direction of an initial feature detection area of the object to be identified in the image to be processed according to an angle and position relation between the connecting line and a coordinate axis of the two-dimensional image coordinate system;

determining the size and the display position of a feature detection area of the object to be identified according to a key point detection result and a connecting line between a first target key point and a second target key point contained in the key point detection result;

and generating a feature detection area corresponding to the object to be identified in the image to be processed according to the size and the display position.

Optionally, the identity recognition method further includes:

determining an overlapping area of a first feature detection area and at least one second feature detection area in the image to be processed, wherein the first feature detection area is one of the at least two feature detection areas in the image to be processed;

and under the condition that the size of the overlapping area is smaller than a preset threshold value, extracting the image features of the first feature detection area.

Optionally, the identity recognition method further includes:

under the condition that the size of the overlapping area is determined to be larger than or equal to a preset threshold value, determining the shielding relation between the first feature detection area and the object to be identified in the at least one second feature detection area;

if the at least one second feature detection area is shielded by the first feature detection area, extracting the image features of the first feature detection area; alternatively, the first and second electrodes may be,

and if the first feature detection area is blocked by the at least one second feature detection area, extracting the image features of the at least one second feature detection area.

Optionally, the identity recognition method further includes:

under the condition that the size of the overlapping area is smaller than a preset threshold value, judging whether the number of the target key points contained in the first feature detection area is larger than or equal to a preset number threshold value or not;

if yes, judging whether the first feature detection area contains at least one target key point of the object to be identified in the at least one second feature detection area;

and if not, extracting the image characteristics of the first characteristic detection area.

Optionally, the identity recognition method further includes:

generating mask information of the feature detection area according to the connection relation of each key point in the feature detection area;

processing the feature detection area according to the mask information to generate a target feature extraction area;

and extracting the image characteristics of the target characteristic detection area, and carrying out pedestrian re-identification on the object to be identified by utilizing the image characteristics to generate an identity identification result corresponding to the object to be identified.

According to a second aspect of embodiments herein, there is provided an identification apparatus comprising:

the detection module is configured to detect key points of an object to be identified contained in the acquired image to be processed;

the determining module is configured to determine a feature detection area corresponding to the object to be identified according to a key point detection result;

the identification module is configured to extract image features of the feature detection area, perform pedestrian re-identification on the object to be identified by using the image features, and generate an identity identification result corresponding to the object to be identified.

According to a third aspect of embodiments herein, there is provided a computing device comprising:

a memory and a processor;

the memory is configured to store computer-executable instructions that, when executed by the processor, perform the steps of the identification method.

According to a fourth aspect of embodiments herein, there is provided a computer-readable storage medium storing computer-executable instructions that, when executed by a processor, perform the steps of the identification method.

In one embodiment of the present description, a key point detection is performed on an object to be recognized included in an acquired image to be processed, a feature detection area corresponding to the object to be recognized is determined according to a key point detection result, image features of the feature detection area are extracted, pedestrian re-recognition is performed on the object to be recognized by using the image features, and an identity recognition result corresponding to the object to be recognized is generated.

In the embodiment of the present specification, the key point detection is performed on the object to be identified in the image to be processed, and the feature detection area of the object to be identified is determined by using the key point position of the object to be identified in the detection result, so as to ensure that the feature detection area contains the features of the object to be identified as much as possible, and contains the features of other objects as little as possible, so as to reduce the influence of the features of other objects on the identification result of the object to be identified, thereby being beneficial to ensuring the accuracy of the feature extraction result of the object to be identified, which is obtained by performing feature extraction on the feature detection area, and further being beneficial to improving the accuracy of the identification result of the object to be identified.

Drawings

FIG. 1 is a flow chart of a process for a method of identification provided in one embodiment of the present description;

FIG. 2 is a diagram illustrating an image to be processed according to an embodiment of the present disclosure;

FIG. 3 is a schematic diagram illustrating an identity recognition method applied to a user trajectory tracking scenario according to an embodiment of the present disclosure;

FIG. 4 is a schematic diagram of an identification device provided in an embodiment of the present disclosure;

fig. 5 is a block diagram of a computing device according to an embodiment of the present disclosure.

Detailed Description

In the following description, numerous specific details are set forth in order to provide a thorough understanding of the present description. This description may be implemented in many ways other than those specifically set forth herein, and those skilled in the art will appreciate that the present description is susceptible to similar generalizations without departing from the scope of the description, and thus is not limited to the specific implementations disclosed below.

The terminology used in the description of the one or more embodiments is for the purpose of describing the particular embodiments only and is not intended to be limiting of the description of the one or more embodiments. As used in one or more embodiments of the present specification and the appended claims, the singular forms "a," "an," and "the" are intended to include the plural forms as well, unless the context clearly indicates otherwise. It should also be understood that the term "and/or" as used in one or more embodiments of the present specification refers to and encompasses any and all possible combinations of one or more of the associated listed items.

It should be understood that although the terms first, second, etc. may be used herein in one or more embodiments to describe various information, these information should not be limited by these terms. These terms are only used to distinguish one type of information from another. For example, a first can also be referred to as a second and, similarly, a second can also be referred to as a first without departing from the scope of one or more embodiments of the present description. The word "if," as used herein, may be interpreted as "at \8230; \8230when" or "when 8230; \823030when" or "in response to a determination," depending on the context.

First, the noun terms referred to in one or more embodiments of the present specification are explained.

Human body key points: the academy corresponds to the position animation. Refers to the obvious key points of bones in the human body. Common in the presently disclosed data sets are 14 points, 16 points, 23 points, and the like.

Pedestrian re-identification (ReID): the technology of identifying the identity of a person by using a 2D human body image is similar to face identification.

In the present specification, an identification method is provided, and the present specification relates to an identification apparatus, a computing device, and a computer-readable storage medium, which are described in detail in the following embodiments one by one.

Fig. 1 shows a process flow diagram of an identification method according to an embodiment of the present disclosure, which includes steps 102 to 106.

And 102, detecting key points of the to-be-identified object contained in the acquired to-be-processed image.

Specifically, the image to be processed is a video frame acquired by a camera, the object to be recognized is an object to be subjected to identity recognition, and the objects to be recognized are different in different identity recognition scenes, for example, in a user trajectory tracking scene, the object to be recognized may be a user; in the scene of pet identification, the object to be identified may be a pet.

Before the identity recognition of the object to be recognized, the key point detection of the object to be recognized can be performed, so that the feature detection area of the object to be recognized is determined according to the key point detection result, and the identity recognition of the object to be recognized can be performed by using the features contained in the feature detection area.

In specific implementation, the key point detection is performed on the object to be identified contained in the acquired image to be processed, and the method can be specifically implemented in the following manner:

the keypoint detection model is trained in the following way:

Specifically, in the embodiments of the present description, a key point detection model may be used to perform key point detection on the object to be identified in the image to be processed, specifically, the image to be processed may be input into the key point detection model, and the key point detection model marks the key point of the user to be identified in the image to be processed.

If the object to be identified is a user, the key point detection model can detect and mark key points of one or more parts of the head, the neck, the shoulders, the crotch, the knees and the feet of the user; if the object to be identified is a pet, the key point detection model can detect and mark key points of one or more parts of the head, ears, nose and feet of the pet; or if the object to be identified is a pet, and the pet is identified by identifying the nasal print characteristics of the pet, the image to be processed is the nasal print image of the pet, and the key point detection model can detect and mark at least one of the feature point where the vertical line of the center of the animal nose in the nasal print image is connected with the upper and lower side boundaries of the nose outline, the feature point of the center of the animal nose, and the feature point where the vertical line of the center of the animal nose is connected with the horizontal line as the key point.

In specific implementation, the scene environment data under different scenes can be used for training the key point detection model so as to train and obtain the key point detection model suitable for different scenes. Specifically, scene images (sample images) under different scenes are obtained, key point labeling is carried out on an object to be identified in the scene images, the sample images are used as training samples, key point labeling results of the sample images are used as labels, a key point detection model to be trained is trained, and a key point detection model is obtained.

In practical applications, the key point detection model may be an HRNet model or other models capable of implementing a key point detection function, and is not limited herein.

The key point detection model is used for detecting the key points of the user to be identified in the image to be processed, so that the detection efficiency is improved, and the accuracy of the key point detection result is improved.

And 104, determining a feature detection area corresponding to the object to be identified according to the key point detection result.

Specifically, after a key point detection result of the object to be recognized is obtained, the feature detection area of the object to be recognized can be determined according to the key point detection result, so that the feature detection area can include features of other objects as less as possible under the condition that the feature detection area includes the features of the object to be recognized as much as possible, and the influence of the features of other objects on the identity recognition result of the object to be recognized is reduced.

As described above, the key point detection model may detect and label key points of one or more portions of the object to be recognized, and therefore, in a case where at least two key points of the object to be recognized are included in a key point detection result generated by the key point detection model performing key point detection on the object to be recognized, two target key points may be selected from the at least two key points, so as to determine the feature detection area of the object to be recognized by using the two target key points.

In an embodiment provided by this specification, determining a feature detection area corresponding to the object to be identified by using two target key points in the key points may specifically be implemented in the following manner:

Specifically, if two target key points are selected from the key point detection results to determine the feature detection area of the object to be identified by using the two target key points, in order to ensure the accuracy of the determination result of the feature detection area, the first target key point and the second target key point that are screened need to satisfy a preset screening condition, for example, a connection line between the first target key point and the second target key point needs to be parallel or perpendicular to a plane (such as the ground) where the object to be identified is located, and may be specifically determined according to actual requirements, which is not limited herein.

Under the condition that a key point detection result generated by the key point detection model through key point detection on an object to be identified comprises at least two key points of the object to be identified, a first target key point and a second target key point can be screened according to a preset screening condition, and after the first target key point and the second target key point are obtained through screening, a feature detection area of the object to be identified can be determined based on a connecting line between the first target key point and the second target key point and the key point detection result.

If the feature detection area may be a rectangular frame, two sides of the feature detection area determined based on the connection line between the first target key point and the second target key point and the key point detection result are perpendicular to the connection line, the other two sides are parallel to the connection line, and the feature detection area includes all key points in the key point detection result.

Further, if the object to be identified is a user, the first target key point may be a head key point of the user, the second target key point may be a middle point of a connecting line of two foot key points of the user, if the feature detection region may be a rectangular frame, the connecting line of the first target key point and the second target key point may be taken as a symmetry axis of the rectangular frame, the middle point of the connecting line of the first target key point and the second target key point may be taken as a central point of the rectangular frame, so as to generate a feature detection region of the user, a length of the feature detection region is greater than a distance of the connecting line of the first target key point and the second target key point, a width of the feature detection region is greater than a distance of the connecting line of the two foot key points of the user, but in actual application, the length and the width of the feature detection region may be determined according to specific requirements, and are not limited herein.

In the embodiment of the present specification, a connection line between a first target key point and a second target key point is taken as a symmetry axis of a rectangular frame, and a midpoint of the connection line between the first target key point and the second target key point is taken as a central point of the rectangular frame, so as to generate a feature detection area of the user, which is beneficial to ensuring accuracy of a determination result of the feature detection area, so that features of an object to be identified are included as much as possible in the feature detection area, and features of other objects are included as little as possible, thereby being beneficial to reducing influence of the features of the other objects on an identification result of the object to be identified, and improving accuracy of the identification result of the object to be identified.

In addition, the feature detection area corresponding to the object to be identified is determined according to the detection result of the key point, and the method can also be realized by the following steps:

calculating a second coordinate of any point on a straight line which passes through the three-dimensional reference coordinate and is parallel to a vertical axis of the three-dimensional reference coordinate system in the image coordinate system according to the three-dimensional reference coordinate and the coordinate conversion relation;

Specifically, under the condition that a key point detection result generated by a key point detection model for detecting a key point of an object to be recognized contains at least two key points of the object to be recognized, whether the key point detection result contains a first target key point or not can be determined, and if the key point detection result contains the first target key point, a feature detection area of the object to be recognized can be determined by using the first target key point.

Specifically, a two-dimensional image coordinate system may be established with any vertex of the image to be processed as an origin of coordinates, a three-dimensional reference coordinate system may be established with any point of a plane (such as the ground) where the object to be recognized is located as an origin, then a first coordinate of the first target key point in the two-dimensional image coordinate system may be determined, and a reference coordinate of the first target key point in the three-dimensional reference coordinate system may be calculated according to a conversion relationship between the two-dimensional image coordinate system and the three-dimensional reference coordinate system, so as to further determine the feature detection area of the object to be recognized by determining a projection of a normal line, which passes through the first target key point in the three-dimensional reference coordinate system and is parallel to a vertical axis, in the two-dimensional image coordinate system.

Further, if the object to be identified is a user, the first target key point may be a midpoint of a connecting line of two foot key points of the user.

In the embodiment of the specification, under the condition that the foot key points of the user in the image to be processed are determined not to be blocked, the feature detection area of the user can be determined by using a normal passing through the middle point, and in practical application, the normal is parallel to a vertical axis in a three-dimensional reference coordinate system and is visually vertical to a plane where the user is located; however, due to the problem of the shooting angle of the camera, the object to be recognized may be visually tilted in the shot image to be processed, and a schematic diagram of the image to be processed provided in the embodiment of the present disclosure is shown in fig. 2, where a user U visually tilts, in this case, in order to ensure that the determined feature detection area can include as many features of the object to be recognized as possible and as few features of other objects as possible, the feature detection area of the object to be recognized also needs to be tilted by a certain angle, and the direction of the tilt and the size of the tilt angle need to be determined according to the projection of the normal in the three-dimensional world coordinate system on the two-dimensional image coordinate system.

Specifically, the coordinate of the middle point of the connecting line of the two foot key points of the user in the two-dimensional image coordinate system is determined, the reference coordinate of the middle point of the connecting line of the two foot key points of the user in the three-dimensional reference coordinate system is determined by utilizing the coordinate conversion relation between the two-dimensional image coordinate system and the three-dimensional reference coordinate system, and as the normal line which passes through the middle point and is perpendicular to the plane where the user is located in the three-dimensional reference coordinate system is parallel to the vertical axis, any point on the normal line (the horizontal and vertical coordinates are equal to those in the reference coordinate of the middle point) can be taken, the coordinate of the point in the two-dimensional image coordinate system is determined, so that the projection position of the point in the image to be processed is determined, and the feature detection area is determined according to the projection position, the middle point and the key points in the key point detection result.

If the coordinate of the middle point of the connecting line of the two foot key points of the user in the two-dimensional image coordinate system is (x) ₀ ,y ₀ ) And assuming that the coordinate of the midpoint in the three-dimensional reference coordinate system is P _w0 (P _xw ,P _yw 0), then both satisfy the following equation:

wherein K is camera internal reference, and R, t are camera external reference.

Will (x) ₀ ,y ₀ ) Substituting the above equation to obtain P _w0 Then in the passage P _w0 And arbitrarily selecting a point P on a normal line parallel to the vertical axis _w1 (P _xw ,P _yw 1), then P is added _w1 Substituting the above equation to obtain P _w1 Coordinates (x) in a two-dimensional image coordinate system ₁ ,y ₁ )。

If the feature detection area may be a rectangular box, the coordinate (x) may be set ₀ ,y ₀ ) And coordinates (x) ₁ ,y ₁ ) The connecting line of the two points is used as a symmetry axis of the rectangular frame, the midpoint of the connecting line of the two points is used as a central point of the rectangular frame to generate a feature detection area of the user, the length of the feature detection area is larger than the distance between the connecting lines of the first target key point and the second target key point, and the width of the feature detection area is larger than the distance between the connecting lines of the two foot key points of the userAnd (5) limiting.

The embodiment of the specification determines the feature detection area according to the connecting line between the first target key point and the coordinate point of the second coordinate and the key point detection result, so that the accuracy of the determination result of the feature detection area is favorably ensured, the feature detection area contains the features of the object to be recognized as much as possible, and contains the features of other objects as little as possible, the influence of the features of other objects on the recognition result of the object to be recognized is favorably reduced, and the accuracy of the recognition result of the object to be recognized is improved.

In specific implementation, the feature detection area corresponding to the object to be identified is determined according to the detection result of the key point, and the method can be realized by the following steps:

if the fact that the first target key point is not contained in the key point detection result is determined, establishing a two-dimensional image coordinate system by taking any vertex of the image to be processed as a coordinate origin;

calculating a second coordinate of any point on a straight line which passes through the reference coordinate and is parallel to the vertical axis of the three-dimensional reference coordinate system in the image coordinate system according to the reference coordinate and the coordinate conversion relation;

Specifically, under the condition that a key point detection result generated by a key point detection model for detecting a key point of an object to be recognized contains at least two key points of the object to be recognized, whether the key point detection result contains a first target key point or a second target key point can be determined, and if the key point detection result does not contain the first target key point but contains the second target key point, a feature detection area of the object to be recognized can be determined by using the second target key point.

Specifically, a two-dimensional image coordinate system may be established with any vertex of the image to be processed as an origin of coordinates, a three-dimensional reference coordinate system may be established with any point of a plane (such as the ground) where the object to be recognized is located as an origin, then a first coordinate of the second target key point in the two-dimensional image coordinate system may be determined, and a reference coordinate of the second target key point in the three-dimensional reference coordinate system may be calculated according to a conversion relationship between the two-dimensional image coordinate system and the three-dimensional reference coordinate system, so as to further determine the feature detection area of the object to be recognized by determining a projection of a normal line, which passes through the first target key point in the three-dimensional reference coordinate system and is parallel to a vertical axis, in the two-dimensional image coordinate system.

Further, if the object to be identified is a user, the first target key point may be a middle point of a connecting line of two foot key points of the user, and the second target key point may be a head key point of the user.

In the embodiment of the specification, when it is determined that the foot key points of the user in the image to be processed are shielded but the head key points are not shielded, the feature detection area of the user can be determined by using the normal line passing through the head key points, and in practical application, the normal line is parallel to the vertical axis in the three-dimensional reference coordinate system and is visually vertical to the plane where the user is located; however, due to the problem of the shooting angle of the camera, the object to be recognized may be visually tilted in the shot image to be processed, in this case, in order to ensure that the determined feature detection area can contain as many features of the object to be recognized as possible and as few features of other objects as possible, the feature detection area of the object to be recognized also needs to be tilted by a certain angle, and the tilting direction and the tilting angle need to be determined according to the projection of the normal line in the three-dimensional world coordinate system in the two-dimensional image coordinate system.

Specifically, the coordinates of the head key point of the user in the two-dimensional image coordinate system are determined, the coordinate conversion relation between the two-dimensional image coordinate system and the three-dimensional reference coordinate system is utilized to determine the reference coordinates of the head key point of the user in the three-dimensional reference coordinate system, and since the normal line which passes through the head key point in the three-dimensional reference coordinate system and is perpendicular to the plane where the user is located is parallel to the vertical axis, any point on the normal line (the horizontal and vertical coordinates are equal to those in the reference coordinates of the head key point) can be taken, and the coordinates of the point in the two-dimensional image coordinate system are determined to determine the projection position of the point in the image to be processed, so that the feature detection area is determined according to the projection position, the middle point and the key point in the key point detection result.

If the head key point coordinate of the user in the two-dimensional image coordinate system is (x) ₀ ,y ₀ ) And assuming that the coordinate of the head key point in the three-dimensional reference coordinate system is P _w0 (P _xw ,P _yw H), then both still satisfy the transformation relationship of equation (1), will (x) ₀ ,y ₀ ) Substituting the above equation to obtain P _w0 Then in the passage of P _w0 And arbitrarily selecting a point P on a normal line parallel to the vertical axis _w1 (P _xw ,P _yw 1), then P is added _w1 Substituting the above equation to obtain P _w1 Coordinates (x) in a two-dimensional image coordinate system ₁ ,y ₁ )。

Wherein h is the average height, and is 170cm in practice.

If the feature detection area is a rectangular frame, the coordinates (x) can be determined ₀ ,y ₀ ) And coordinates (x) ₁ ,y ₁ ) The connection line of the two points is used as the symmetry axis of the rectangular frame, the midpoint of the connection line of the two points is used as the center point of the rectangular frame to generate the feature detection area of the user, the length of the feature detection area is larger than the distance between the connection lines of the first target key point and the second target key point, and the width of the feature detection area is larger than the distance between the connection lines of the two foot key points of the user.

The embodiment of the specification determines the feature detection area according to the connecting line between the first target key point and the coordinate point of the second coordinate and the key point detection result, so that the accuracy of the determination result of the feature detection area is favorably ensured, the feature detection area contains the features of the object to be identified as much as possible, and contains the features of other objects as little as possible, the influence of the features of other objects on the identification result of the object to be identified is favorably reduced, and the accuracy of the identification result of the object to be identified is improved.

In addition, the feature detection area corresponding to the object to be identified is determined according to the key point detection result, and the method can also be realized by the following mode:

determining whether the key point detection result comprises a first target key point and a second target key point;

Specifically, under the condition that a key point detection result generated by the key point detection model for detecting the key point of the object to be recognized includes at least two key points of the object to be recognized, the first target key point and the second target key point may be screened according to a preset screening condition, and after the first target key point and the second target key point are obtained by screening, the feature detection area of the object to be recognized may be determined based on a connection line between the first target key point and the second target key point and the key point detection result.

The feature detection area may be a rectangular frame, and an initial feature detection area of the object to be processed may be determined before the object to be recognized in the image to be processed is subjected to the keypoint detection, where the initial feature detection area is not visually tilted, and a specific schematic diagram is shown in fig. 2. However, due to the problem of the shooting angle of the camera, the object to be recognized may be visually tilted in the shot image to be processed, and in this case, in order to ensure that the determined feature detection area can include as many features of the object to be recognized as possible and as few features of other objects as possible, the initial feature detection area needs to be rotated according to a certain tilting direction and a certain tilting angle, where the tilting direction and the tilting angle need to be determined according to a tilting angle of a connecting line between a first target key point and a second target key point of the object to be processed.

In practical application, a two-dimensional image coordinate system can be established based on any vertex of the image to be processed as an origin, and the coordinates (x) of the first target key point and the second target key point in the two-dimensional image coordinate system are determined ₀ ,y ₀ ) And coordinates (x) ₁ ,y ₁ ) Determining the rotation angle and the rotation direction of the initial characteristic detection area according to the angle and the position relation between the connecting line of the two coordinate points and the edge of the image to be processed; alternatively, the coordinates (x) may be divided ₀ ,y ₀ ) And coordinates (x) ₁ ,y ₁ ) And substituting the formula (2) into the coordinate system to determine the inclination angle of the connecting line between the first target key point and the second target key point in the two-dimensional coordinate system.

And after the inclination angle is obtained through calculation, rotating the initial feature detection area according to the size and the inclination direction of the inclination angle, and thus obtaining the feature detection area of the object to be identified.

When the rotation angle and the rotation direction are known, the center point position of the initial feature detection area is kept unchanged, the initial feature detection area is rotated according to the rotation angle and the rotation direction, the rotation frame is appropriately enlarged to be used as the feature detection area of the object to be recognized, and a schematic diagram of the rotation result is shown in fig. 2. Wherein, the expansion strategy may be: the body width is not less than 0.3 times the body length in consideration of the aspect ratio of the object to be recognized, and the body length is enlarged by 10% vertically and 20% horizontally.

The initial feature detection area is rotated to obtain the feature detection area, so that the accuracy of the feature extraction result of the object to be recognized, which is obtained by extracting the features of the feature detection area, is favorably ensured, and the accuracy of the recognition result of the object to be recognized is further favorably improved.

Specifically, under the condition that a key point detection result generated by the key point detection model through key point detection on an object to be recognized contains at least two key points of the object to be recognized, whether the key point detection result contains a first target key point or not can be determined, and if the key point detection result contains the first target key point, a feature detection area of the object to be recognized can be determined by using the first target key point.

Specifically, a two-dimensional image coordinate system may be established with any vertex of the image to be processed as an origin of coordinates, a three-dimensional reference coordinate system may be established with any point of a plane (such as the ground) where the object to be recognized is located as an origin, then a first coordinate of the first target keypoint in the two-dimensional image coordinate system may be determined, and a reference coordinate of the first target keypoint in the three-dimensional reference coordinate system may be calculated according to a conversion relationship between the two-dimensional image coordinate system and the three-dimensional reference coordinate system, so as to further determine the rotation angle and the rotation direction of the initial feature detection area by determining a projection of a normal line, which passes through the first target keypoint in the three-dimensional reference coordinate system and is parallel to the vertical axis, in the two-dimensional image coordinate system.

In the embodiment of the description, when it is determined that the key point of the foot of the user in the image to be processed is not blocked, the projection position of any point on the normal line passing through the midpoint in the two-dimensional image coordinate system can be used for determining the rotation angle and the rotation direction of the initial feature detection area, and in practical application, the normal line is parallel to the vertical axis in the three-dimensional reference coordinate system and is visually vertical to the plane where the user is located; however, due to the problem of the shooting angle of the camera, the object to be recognized may be visually tilted in the shot image to be processed, and in this case, in order to ensure that the determined feature detection area can contain as many features of the object to be recognized as possible and as few features of other objects as possible, the initial feature detection area needs to be rotated according to a certain tilting direction and a certain tilting angle, wherein the tilting direction and the tilting angle need to be determined according to the projection of the normal line in the three-dimensional world coordinate system on the two-dimensional image coordinate system.

If the coordinate of the middle point of the connecting line of the two foot key points of the user in the two-dimensional image coordinate system is (x) ₀ ,y ₀ ) And assuming that the coordinate of the midpoint in the three-dimensional reference coordinate system is P _w0 (P _xw ,P _yw 0), then the coordinates (x) can be calculated ₀ ,y ₀ ) Substituting into formula (1) to obtain P _w0 Then in the passage of P _w0 And arbitrarily selecting a point P on a normal line parallel to the vertical axis _w1 (P _xw ,P _yw 1), then P is added _w1 Substituting the formula (1) to obtain P _w1 Coordinates (x) in a two-dimensional image coordinate system ₁ ,y ₁ )。

Then, according to the angle and position relation between the connecting line of the two coordinate points and the edge of the image to be processed, determining the rotation angle and the rotation direction of the initial characteristic detection area; alternatively, the coordinates (x) may be expressed ₀ ,y ₀ ) And coordinates (x) ₁ ,y ₁ ) And substituting the formula (2) into the formula (2) to determine the inclination angle of the connecting line between the first target key point and the second target key point in the two-dimensional coordinate system.

The initial feature detection area is rotated to obtain the feature detection area, so that the accuracy of the feature extraction result of the object to be recognized, which is obtained by extracting the features of the feature detection area, is ensured, and the accuracy of the recognition result of the object to be recognized is further improved.

In addition, the feature detection area corresponding to the object to be identified is determined according to the detection result of the key point, and the method can also be realized through the following steps:

Specifically, if two target key points are selected from the key point detection results to determine the feature detection area of the object to be identified by using the two target key points, in order to ensure the accuracy of the determination result of the feature detection area, the first target key point and the second target key point that are screened need to satisfy a preset screening condition, for example, a connection line between the first target key point and the second target key point needs to be parallel or perpendicular to a plane (such as the ground) where the object to be identified is located, and the first target key point and the second target key point can be specifically determined according to actual requirements without limitation.

Under the condition that a key point detection result generated by the key point detection model for detecting the key points of the object to be recognized contains at least two key points of the object to be recognized, the first target key point and the second target key point can be screened according to a preset screening condition, and after the first target key point and the second target key point are obtained through screening, the feature detection area of the object to be recognized can be determined based on a connecting line between the first target key point and the second target key point and the key point detection result.

Further, if the object to be identified is a user, the first target key point may be a head key point of the user, the second target key point may be a middle point of a connection line of two foot key points of the user, if the feature detection region may be a rectangular frame, the connection line of the first target key point and the second target key point may be taken as a symmetry axis of the rectangular frame, the middle point of the connection line of the first target key point and the second target key point may be taken as a central point of the rectangular frame, so as to generate a feature detection region of the user, a length of the feature detection region is greater than a distance of the connection line of the first target key point and the second target key point, a width of the feature detection region is greater than a distance of the connection line of the two foot key points of the user, and a specific determination policy of the size of the feature detection region may be: the body width is 10% enlarged vertically and 20% enlarged horizontally in the body length direction, and the body width is not less than 0.3 times the body length in consideration of the aspect ratio of the object to be recognized.

In practical applications, the length and width of the feature detection area may be determined according to specific requirements, and are not limited herein.

The embodiment of the specification determines the feature detection area of the object to be recognized by using the key point position of the object to be recognized in the detection result, so as to ensure that the feature detection area contains the features of the object to be recognized as much as possible and contains the features of other objects as little as possible, so as to reduce the influence of the features of other objects on the recognition result of the object to be recognized, thereby being beneficial to improving the accuracy of the recognition result of the object to be recognized.

And 106, extracting the image characteristics of the characteristic detection area, and performing pedestrian re-identification on the object to be identified by using the image characteristics to generate an identity identification result corresponding to the object to be identified.

Specifically, after the feature detection area of the object to be recognized is determined, the image features in the feature detection area can be extracted, and the object to be recognized is subjected to pedestrian re-recognition based on the extracted image features, so that whether a target user exists in the image to be processed is determined, and the trajectory tracking of the object to be recognized is realized.

In addition, after the feature detection area of the object to be recognized is determined, the feature detection area can be screened, which can be specifically realized by the following method:

Further, under the condition that the size of the overlapping area is determined to be larger than or equal to a preset threshold value, determining the shielding relation between the first feature detection area and the object to be identified in the at least one second feature detection area;

Specifically, after a feature detection region of an object to be identified is determined, the feature detection region is screened in a manner that an overlapping area of a first feature detection region and at least one second feature detection region is determined, and under the condition that the overlapping area is determined to be smaller than a preset threshold value, image features of the first feature detection region can be extracted.

In practical application, namely whether aggregation shielding exists between the feature detection areas is judged, and the feature detection areas without aggregation shielding are directly used for extracting image features; the criterion for judging whether the aggregation occlusion exists is as follows: the area of the overlapping part of any other characteristic detection area and the current characteristic detection area in the image to be processed accounts for more than 20% of the area of the current characteristic detection area. And judging the shielding relation among a plurality of feature detection areas for the feature detection areas with the gathered shielding to determine whether the feature detection areas are foreground or shielded, wherein the feature detection areas can be used for extracting image features if the feature detection areas are foreground, and the feature detection areas are discarded if the feature detection areas are shielded.

By screening the feature detection area, the accuracy of the feature extraction result of the object to be recognized, which is obtained by extracting the features of the screened feature detection area, is ensured, and the accuracy of the recognition result of the object to be recognized is further improved.

In addition, under the condition that the size of the overlapping area is smaller than a preset threshold value, whether the number of the target key points contained in the first feature detection area is larger than or equal to a preset number threshold value or not is judged;

if so, judging whether the first feature detection area contains at least one target key point of the object to be identified in the at least one second feature detection area;

Specifically, considering that in practical application, an article may block key points, in order to ensure accuracy of an identification result obtained by performing identity identification on an object to be identified by using remaining key points under the condition that part of key points of the object to be identified are blocked, the embodiment of the present specification may determine a blocking relationship by using 6 key points of the head, the neck, the left and right shoulders, and the left and right thighs of the object to be identified (user). And satisfies, in the feature detection area: when key points of the head and the neck are visible, key points of the left shoulder and the left crotch are visible simultaneously or key points of the right shoulder and the right crotch are visible simultaneously, the area of the overlapping part of the other feature detection region and the current feature detection region accounts for less than 0.8 of the area of the current feature detection region, and does not contain key points of the head, the neck and any other part in the other feature detection region, or does not contain key points of the neck, the shoulder and the crotch on any side in the other feature detection region, the feature detection region can be used for extracting image features, otherwise, the feature detection region is not usable.

In addition, after the feature detection area corresponding to the object to be identified is determined according to the key point detection result, the target feature extraction area in each feature detection area can be determined according to the connection relationship of each key point in the feature detection area, which can be specifically realized by the following method:

Specifically, in the embodiments of the present specification, an image of an object to be recognized in a feature detection area is processed by using a connection relationship between a key point and a key point in the feature detection area, so as to generate mask information of the object to be recognized, and to weaken, by using the mask information, an influence of another feature on a feature of the object to be recognized.

Under the condition that the object to be identified is a user, the specific implementation steps of the process are as follows:

sequentially connecting head, neck, shoulders and double-span points, connecting corresponding elbow and hand key points of the key points for visible shoulder key points, and connecting key points of the crotch and the knee and foot for visible crotch;

for all the connecting lines, generating masks around the lines according to Gaussian distribution, wherein the variance of the Gaussian distribution is 1/2 of the size of the short side of the feature detection area (rectangular frame);

after multiplying the original feature extraction area by mask information, generating a target feature extraction area, wherein the target feature extraction area is displayed in a heat map form, and the whiter the color of the target feature extraction area is, the higher the corresponding weight is;

after the target feature extraction area is generated, the image features of the target feature detection area can be extracted, the image features are utilized to carry out pedestrian re-identification on the object to be identified, and an identity identification result corresponding to the object to be identified is generated.

In practical application, the image features can be extracted by using a neural network.

In addition, if the characteristic identification result of the object to be identified can be identified and obtained, the image characteristics of the object to be identified can be put in a warehouse, namely, stored in a database; if the characteristic recognition result of the object to be recognized cannot be obtained through recognition, the identity information of the object to be recognized needs to be inquired from a characteristic library, after the identity information of the object to be recognized is obtained through inquiry, the image characteristics of the object to be recognized are put into a storage, and a sample set for training the neural network is expanded based on the image characteristics.

In the embodiment of the specification, the key point detection is performed on the object to be recognized in the image to be processed, and the key point position of the object to be recognized in the detection result is used for determining the feature detection area of the object to be recognized, so that the feature detection area is ensured to contain the features of the object to be recognized as much as possible, and the features of other objects are reduced as little as possible, so that the influence of the features of other objects on the recognition result of the object to be recognized is reduced, the accuracy of the feature extraction result of the object to be recognized, which is obtained by extracting the features of the feature detection area, is favorably ensured, and the accuracy of the recognition result of the object to be recognized is further favorably improved.

The following description will further describe the identity recognition method with reference to fig. 3 by taking an application of the identity recognition method provided in this specification in a user trajectory tracking scenario as an example. Fig. 3 is a schematic diagram illustrating an application of an identity recognition method to a user trajectory tracking scenario according to an embodiment of the present specification.

Firstly, carrying out human body detection, namely carrying out human body detection on an image to be processed to detect the position of a human body in the image to be processed, determining an initial rectangular frame corresponding to the human body, further detecting the key point position of the human body to carry out shielding prediction on the human body according to the key point position, specifically judging the shielding relation of the human body by using 6 points of the head, the neck, the left shoulder, the right shoulder and the left crotch, and continuing the subsequent step of correcting the rectangular rotating frame if the 6 key points are not shielded;

in practical application, a two-dimensional image coordinate system is established in an image to be processed, coordinates of a head key point of a user and a middle point of a connecting line of two foot key points in the two-dimensional image coordinate system are determined, an inclination angle of the connecting line between the head key point and the middle point is determined by using the two coordinates, and after the inclination angle is obtained through calculation, the initial feature detection area is rotated according to the size and the inclination direction of the inclination angle, so that a feature detection area of the object to be identified can be obtained;

after the feature detection area is obtained, foreground person screening can be performed, specifically, if the area of the overlapping part of any other feature detection area and the current feature detection area in the image to be processed accounts for more than 20% of the area of the current feature detection area, the shielding relation between the other feature detection area and the current feature detection area needs to be judged to determine whether the current feature detection area is a foreground or a shielded area, if the current feature detection area is the foreground, the image feature can be extracted, and if the current feature detection area is the shielded area, the image feature is discarded;

if the current feature detection area is a foreground, sequentially connecting the head, the neck, the shoulders and the two crotch points of the user, connecting corresponding elbow and hand key points of the key points for visible shoulder key points, and connecting the crotch, the knee and foot key points for visible crotch;

after multiplying the original feature extraction area by mask information, generating a target feature extraction area, wherein the target feature extraction area is displayed in a heat map mode, and the more white the color of the target feature extraction area is, the higher the corresponding weight is;

after the target feature extraction area is generated, the image features of the target feature detection area can be extracted by using a neural network, the pedestrian re-identification is carried out on the object to be identified by using the image features, and an identity identification result corresponding to the object to be identified is generated so as to determine whether a target user exists in the image to be processed, so that the track tracking of the object to be identified is realized.

In addition, if the characteristic identification result of the object to be identified can be identified and obtained, the image characteristics of the object to be identified can be put into a database, namely, the image characteristics are stored into a database; if the characteristic recognition result of the object to be recognized cannot be obtained through recognition, the identity information of the object to be recognized needs to be inquired from a characteristic library, after the identity information of the object to be recognized is obtained through inquiry, the image characteristics of the object to be recognized are put into a warehouse, and a sample set of the neural network is trained through expansion based on the image characteristics.

Corresponding to the above method embodiment, the present specification further provides an identity recognition apparatus embodiment, and fig. 4 shows a schematic diagram of an identity recognition apparatus provided in an embodiment of the present specification. As shown in fig. 4, the apparatus includes:

a detection module 402 configured to perform key point detection on an object to be identified contained in the acquired image to be processed;

a determining module 404 configured to determine a feature detection area corresponding to the object to be identified according to a key point detection result;

the identification module 406 is configured to extract image features of the feature detection area, perform pedestrian re-identification on the object to be identified by using the image features, and generate an identity identification result corresponding to the object to be identified.

Optionally, the detecting module 402 includes:

the detection submodule is configured to input the acquired image to be processed into a key point detection model, wherein the key point detection model is used for performing key point detection on an object to be identified contained in the image to be processed;

the key point detection model is trained in the following way:

and training the key point detection model by using the sample image as a training sample and the key point labeling result as a label to obtain the key point detection model.

Optionally, the determining module 404 includes:

the first determining submodule is configured to determine whether the key point detection result contains a first target key point and a second target key point;

if the operation result of the determination submodule is yes, operating a second determination submodule;

the second determining submodule is configured to determine a feature detection area of the object to be identified based on a connecting line between the first target key point and the second target key point and the key point detection result.

Optionally, the determining module 404 includes:

the first establishing submodule is configured to establish a two-dimensional image coordinate system by taking any vertex of the image to be processed as a coordinate origin if the fact that the key point detection result contains a first target key point is determined;

a third determining submodule configured to determine first coordinates of the first target keypoint in the two-dimensional image coordinate system;

the first calculation sub-module is configured to calculate reference coordinates of the first target key point in a three-dimensional reference coordinate system according to a coordinate conversion relation between a two-dimensional image coordinate system and the three-dimensional reference coordinate system and the first coordinates;

the second calculation submodule is configured to calculate a second coordinate of any point on a straight line which passes through the reference coordinate and is parallel to a vertical axis of the three-dimensional reference coordinate system in the image coordinate system according to the reference coordinate and the coordinate conversion relation;

and the fourth determining submodule is configured to determine the feature detection area of the object to be identified based on a connecting line between the first target key point and the coordinate point of the second coordinate and the key point detection result.

Optionally, the determining module 404 includes:

the second establishing submodule is configured to establish a two-dimensional image coordinate system by taking any vertex of the image to be processed as a coordinate origin if the fact that the first target key point is not contained in the key point detection result is determined;

a fifth determining submodule configured to determine a first coordinate of a second target keypoint in the two-dimensional image coordinate system;

the third calculation sub-module is configured to calculate a reference coordinate of the second target key point in the three-dimensional reference coordinate system according to a coordinate conversion relation between a two-dimensional image coordinate system and a three-dimensional reference coordinate system, the first coordinate and a preset value corresponding to the second target key point in the vertical axis direction of the three-dimensional reference coordinate system;

the fourth calculation submodule is configured to calculate a second coordinate of any point on a straight line which passes through the reference coordinate and is parallel to a vertical axis of the three-dimensional reference coordinate system in the image coordinate system according to the reference coordinate and the coordinate conversion relation;

and the sixth determining submodule is configured to determine the feature detection area of the object to be identified based on a connecting line between the second target key point and a coordinate point of the second coordinate and the key point detection result.

Optionally, the determining module 404 includes:

a seventh determining submodule configured to determine whether the key point detection result includes the first target key point and the second target key point;

an eighth determining submodule configured to determine, if yes, a connection line between the first target key point and the second target key point, and determine a rotation angle and a rotation direction of an initial feature detection area of the object to be identified in the image to be processed according to an angle and a position relationship between the connection line and an edge of the image to be processed;

and the first rotation submodule is configured to rotate the initial feature detection area according to the rotation angle and the rotation direction, and generate a feature detection area corresponding to the object to be identified.

Optionally, the determining module 404 includes:

the third establishing submodule is configured to establish a two-dimensional image coordinate system by taking any vertex of the image to be processed as a coordinate origin if the fact that the detection result of the key point contains the first target key point is determined;

a ninth determining submodule configured to determine first coordinates of the first target keypoint in the two-dimensional image coordinate system;

a fifth calculation submodule configured to calculate a reference coordinate of the first target keypoint in a three-dimensional reference coordinate system according to a coordinate conversion relationship between a two-dimensional image coordinate system and the three-dimensional reference coordinate system and the first coordinate;

a sixth calculating submodule configured to calculate a second coordinate of any one point on a straight line passing through the reference coordinate and being parallel to a vertical axis of the three-dimensional reference coordinate system in the image coordinate system, according to the reference coordinate and the coordinate conversion relationship;

a tenth determining submodule configured to determine a connection line between the first target key point and a coordinate point of the second coordinate, and determine a rotation angle and a rotation direction of an initial feature detection area of the object to be identified in the image to be processed according to an angle and a position relationship between the connection line and a coordinate axis of the two-dimensional image coordinate system;

and the second rotation submodule is configured to rotate the initial feature detection area according to the rotation angle and the rotation direction, and generate a feature detection area corresponding to the object to be identified.

Optionally, the determining module 404 includes:

the position determining submodule is configured to determine the size and the display position of a feature detection area of the object to be identified according to a key point detection result and a connecting line between a first target key point and a second target key point contained in the key point detection result;

and the generation submodule is configured to generate a feature detection area corresponding to the object to be identified in the image to be processed according to the size and the display position.

Optionally, the identity recognition apparatus further includes:

an overlapping area determination module configured to determine an overlapping area of a first feature detection region and at least one second feature detection region in the image to be processed, wherein the first feature detection region is one of the at least two feature detection regions in the image to be processed;

a first extraction module configured to extract an image feature of the first feature detection region if it is determined that the size of the overlapping area is smaller than a preset threshold.

Optionally, the identity recognition apparatus further includes:

an occlusion relation determining module configured to determine an occlusion relation between the first feature detection region and the object to be identified in the at least one second feature detection region if it is determined that the size of the overlapping area is greater than or equal to a preset threshold;

a second extraction module configured to extract image features of the first feature detection area if the at least one second feature detection area is occluded by the first feature detection area; alternatively, the first and second liquid crystal display panels may be,

a third extraction module configured to extract an image feature of the at least one second feature detection area if the first feature detection area is occluded by the at least one second feature detection area.

Optionally, the identity recognition apparatus further includes:

the first judging module is configured to judge whether the number of the target key points contained in the first feature detection area is greater than or equal to a preset number threshold or not under the condition that the size of the overlapping area is smaller than a preset threshold;

if the operation result of the first judgment module is yes, operating a second judgment module;

the second judging module is configured to judge whether the first feature detection area contains at least one target key point of an object to be identified in the at least one second feature detection area;

if the operation result of the second judgment module is negative, operating a fourth extraction module;

the fourth extraction module is configured to extract an image feature of the first feature detection area.

Optionally, the identity recognition apparatus further includes:

the mask information generating module is configured to generate mask information of the feature detection area according to the connection relation of each key point in the feature detection area;

the processing module is configured to process the feature detection area according to the mask information to generate a target feature extraction area;

the identification module is configured to extract image features of the target feature detection area, perform pedestrian re-identification on the object to be identified by using the image features, and generate an identity identification result corresponding to the object to be identified.

The above is a schematic scheme of an identification apparatus of this embodiment. It should be noted that the technical solution of the identity recognition apparatus and the technical solution of the identity recognition method belong to the same concept, and details of the technical solution of the identity recognition apparatus, which are not described in detail, can be referred to the description of the technical solution of the identity recognition method.

FIG. 5 illustrates a block diagram of a computing device 500, provided in accordance with one embodiment of the present specification. The components of the computing device 500 include, but are not limited to, a memory 510 and a processor 520. Processor 520 is coupled to memory 510 via bus 530, and database 550 is used to store data.

Computing device 500 also includes access device 540, access device 540 enabling computing device 500 to communicate via one or more networks 560. Examples of such networks include the Public Switched Telephone Network (PSTN), a Local Area Network (LAN), a Wide Area Network (WAN), a Personal Area Network (PAN), or a combination of communication networks such as the internet. The access device 540 may include one or more of any type of network interface, e.g., a Network Interface Card (NIC), wired or wireless, such as an IEEE802.11 Wireless Local Area Network (WLAN) wireless interface, a worldwide interoperability for microwave access (Wi-MAX) interface, an ethernet interface, a Universal Serial Bus (USB) interface, a cellular network interface, a bluetooth interface, a Near Field Communication (NFC) interface, and so forth.

In one embodiment of the present description, the above-described components of computing device 500, as well as other components not shown in FIG. 5, may also be connected to each other, such as by a bus. It should be understood that the block diagram of the computing device architecture shown in FIG. 5 is for purposes of example only and is not limiting as to the scope of the present description. Other components may be added or replaced as desired by those skilled in the art.

Computing device 500 may be any type of stationary or mobile computing device, including a mobile computer or mobile computing device (e.g., tablet, personal digital assistant, laptop, notebook, netbook, etc.), mobile phone (e.g., smartphone), wearable computing device (e.g., smartwatch, smartglasses, etc.), or other type of mobile device, or a stationary computing device such as a desktop computer or PC. Computing device 500 may also be a mobile or stationary server.

Wherein the memory 510 is configured to store computer-executable instructions and the processor 520 is configured to execute the following computer-executable instructions:

determining a characteristic detection area corresponding to the object to be identified according to a key point detection result;

The foregoing is a schematic diagram of a computing device of the present embodiment. It should be noted that the technical solution of the computing device and the technical solution of the above-mentioned identity recognition method belong to the same concept, and details that are not described in detail in the technical solution of the computing device can be referred to the description of the technical solution of the above-mentioned identity recognition method.

An embodiment of the present specification also provides a computer readable storage medium storing computer instructions, which when executed by a processor, are used for implementing the steps of the identity recognition method.

The above is an illustrative scheme of a computer-readable storage medium of the present embodiment. It should be noted that the technical solution of the storage medium belongs to the same concept as the technical solution of the above-mentioned identification method, and details that are not described in detail in the technical solution of the storage medium can be referred to the description of the technical solution of the above-mentioned identification method.

The foregoing description of specific embodiments has been presented for purposes of illustration and description. Other embodiments are within the scope of the following claims. In some cases, the actions or steps recited in the claims may be performed in a different order than in the embodiments and still achieve desirable results. In addition, the processes depicted in the accompanying figures do not necessarily require the particular order shown, or sequential order, to achieve desirable results. In some embodiments, multitasking and parallel processing may also be possible or may be advantageous.

The computer instructions comprise computer program code which may be in source code form, object code form, an executable file or some intermediate form, or the like. The computer-readable medium may include: any entity or device capable of carrying the computer program code, recording medium, U.S. disk, removable hard disk, magnetic diskette, optical disk, computer Memory, read-Only Memory (ROM), random Access Memory (RAM), electrical carrier wave signal, telecommunications signal, and software distribution medium, etc. It should be noted that the computer-readable medium may contain suitable additions or subtractions depending on the requirements of legislation and patent practice in jurisdictions, for example, in some jurisdictions, computer-readable media may not include electrical carrier signals or telecommunication signals in accordance with legislation and patent practice.

It should be noted that, for the sake of simplicity, the foregoing method embodiments are described as a series of combinations of acts, but it should be understood by those skilled in the art that the embodiments are not limited by the described order of acts, as some steps may be performed in other orders or simultaneously according to the embodiments. Further, those skilled in the art should also appreciate that the embodiments described in this specification are preferred embodiments and that acts and modules referred to are not necessarily required for an embodiment of the specification.

In the foregoing embodiments, the descriptions of the respective embodiments have respective emphasis, and for parts that are not described in detail in a certain embodiment, reference may be made to the related descriptions of other embodiments.

The preferred embodiments of the present specification disclosed above are intended only to aid in the description of the specification. Alternative embodiments are not exhaustive and do not limit the invention to the precise embodiments described. Obviously, many modifications and variations are possible in light of the above teaching. The embodiments were chosen and described in order to best explain the principles of the embodiments and the practical application, and to thereby enable others skilled in the art to best understand the specification and utilize the specification. The specification is limited only by the claims and their full scope and equivalents.

Claims

1. An identity recognition method comprising:

extracting image features of the feature detection area, and performing pedestrian re-identification on the object to be identified by using the image features to generate an identity identification result corresponding to the object to be identified;

determining a feature detection area corresponding to the object to be identified according to the key point detection result, wherein the determining includes:

2. The identity recognition method according to claim 1, wherein the performing of the key point detection on the object to be recognized included in the acquired image to be processed comprises:

the key point detection model is trained in the following way:

3. The identity recognition method according to claim 1, wherein the determining the feature detection area corresponding to the object to be recognized according to the key point detection result comprises:

4. The identity recognition method according to claim 1 or 3, wherein the determining the feature detection area corresponding to the object to be recognized according to the key point detection result comprises:

calculating a second coordinate of any point on a straight line which passes through the reference coordinate and is parallel to the vertical axis of the three-dimensional reference coordinate system in the two-dimensional image coordinate system according to the reference coordinate and the coordinate conversion relation;

5. The identity recognition method according to claim 1 or 3, wherein the determining the feature detection area corresponding to the object to be recognized according to the key point detection result comprises:

calculating a second coordinate of any point on a straight line which passes through the reference coordinate and is parallel to a vertical axis of the three-dimensional reference coordinate system in the two-dimensional image coordinate system according to the reference coordinate and the coordinate conversion relation;

6. The identity recognition method according to claim 1, wherein the determining the feature detection area corresponding to the object to be recognized according to the key point detection result comprises:

if the first target key point is contained in the key point detection result, establishing a two-dimensional image coordinate system by taking any vertex of the image to be processed as a coordinate origin;

7. The identity recognition method according to claim 1, wherein the determining the feature detection area corresponding to the object to be recognized according to the key point detection result comprises:

8. The identity recognition method of claim 1, further comprising:

9. The identification method of claim 8, further comprising:

10. The identification method of claim 8, further comprising:

11. The identification method of claim 1, further comprising:

and extracting the image characteristics of the target characteristic extraction area, and performing pedestrian re-identification on the object to be identified by using the image characteristics to generate an identity identification result corresponding to the object to be identified.

12. An identification device comprising:

the detection module is configured to detect key points of the to-be-identified object contained in the acquired to-be-processed image;

the recognition module is configured to extract image features of the feature detection area, perform pedestrian re-recognition on the object to be recognized by using the image features, and generate an identity recognition result corresponding to the object to be recognized;

wherein the determining module comprises:

the seventh determining submodule is configured to determine whether the key point detection result contains the first target key point and the second target key point;

13. A computing device, comprising:

a memory and a processor;

the memory is configured to store computer-executable instructions, and the processor is configured to implement the steps of the identification method according to any one of claims 1 to 11 when executing the computer-executable instructions.

14. A computer readable storage medium storing computer instructions which, when executed by a processor, carry out the steps of the identification method of any one of claims 1 to 11.