Pedestrian identity recognition method and device, electronic equipment and readable storage medium

Abstract

The invention provides a pedestrian identity recognition method and device, electronic equipment and a readable storage medium. The method comprises the following steps: acquiring a figure appearance image of a to-be-recognized person; based on a pre-stored personnel archive library, performing matching to obtain a target face image corresponding to the character appearance image; wherein the personnel archive library comprises face images of different personnel and a plurality of candidate person and object appearance images corresponding to the face images; and recognizing the target face image to obtain an identity recognition result of the to-be-recognized person. According to the method, the high accuracy of face recognition and the passive characteristic of figure appearance recognition are fully utilized, the association between the face image and the figure appearance image is established, the identity recognition result of the to-be-recognized person can be obtained under the condition that the figure appearance image of the to-be-recognized person is obtained, cooperation of the recognized person is not needed in the process, and the recognition efficiency is improved. And meanwhile, the accuracy of personnel identity recognition is improved.

Description

Technical Field

[0001] The present invention relates to the field of electronic information technology, and in particular to a pedestrian identification method, device, electronic device and readable storage medium. Background Art

[0002] Personnel identification is a fundamental capability of many artificial intelligence (AI) application systems and a crucial component of building smart cities. Among existing personnel identification technologies, active identification requires the cooperation of the person being identified, making it unsuitable for scenarios such as security, traffic statistics, and intelligent person search. Passive identification, on the other hand, suffers from low accuracy due to posture errors caused by lack of cooperation.

[0003] Personnel identity recognition is the basic capability of many artificial intelligence application systems and an important part of building smart cities. It has a wide range of applications in smart payment, security, and missing persons.

[0004] Currently, facial recognition-based identity verification methods include active and passive methods. Active identification systems are relatively well-established, but they require the cooperation of the person being identified, making them unsuitable for scenarios like security, traffic statistics, and intelligent person search. While passive identification doesn't require the person's cooperation, it often suffers from low accuracy due to issues like image blur and changes in posture and background.

[0005] Therefore, it is very necessary to solve the problem of low recognition accuracy of existing passive identity recognition methods. Summary of the Invention

[0006] The present invention provides a pedestrian identification method, device, electronic device and readable storage medium to overcome the defect of low recognition accuracy of existing passive identification methods. It does not require the cooperation of the person being identified and can achieve high person identification accuracy. The method can then be used to build an efficient and intelligent urban security system, expanding the application scope of person identification.

[0007] On the one hand, the present invention provides a pedestrian identification method, comprising: obtaining a person appearance image of a person to be identified; matching a target facial image corresponding to the person appearance image based on a pre-stored personnel archive; wherein the personnel archive includes facial images of different persons and their corresponding multiple candidate person appearance images; identifying the target facial image to obtain an identity identification result of the person to be identified.

[0008] Furthermore, constructing the personnel archive specifically includes: determining a target area and collecting surveillance videos of the target area during a set time period; extracting facial images corresponding to key frames in the surveillance video based on a face detection algorithm; extracting candidate person appearance images corresponding to key frames in the surveillance video based on a pedestrian detection algorithm; matching the facial images of the same person with the candidate person appearance images, and binding the identity information of the person to construct the personnel archive.

[0009] Furthermore, the matching based on the pre-stored personnel archive to obtain the target facial image corresponding to the person appearance image includes: extracting the first person appearance feature of the person appearance image; extracting the second person appearance feature corresponding to each candidate person appearance image in the personnel archive; calculating the cosine similarity between the first person appearance feature and each of the second person appearance features; and when the cosine similarity is higher than a set threshold, obtaining the facial image corresponding to the candidate person appearance image corresponding to the cosine similarity to obtain the target facial image.

[0010] Furthermore, matching the facial image of the same person with the appearance image of the candidate character includes: establishing a matching relationship between the facial image and the appearance image of the candidate character when the facial image is completely contained in the appearance image of the candidate character.

[0011] Furthermore, the matching of the facial image and the candidate character appearance image of the same person also includes: when the facial image contains candidate character appearance images corresponding to multiple persons, canceling the matching relationship between the facial image and the candidate character appearance image; and / or, when the candidate character appearance image contains facial images corresponding to multiple persons, canceling the matching relationship between the facial image and the candidate character appearance image.

[0012] Furthermore, the person appearance image is a full-body front image, a full-body back image, a partial front image or a partial back image of the person to be identified.

[0013] In a second aspect, the present invention also provides a pedestrian identification device, comprising: an image acquisition module for acquiring a person appearance image of a person to be identified; an image matching module for matching a target facial image corresponding to the person appearance image based on a pre-stored personnel archive; wherein the personnel archive includes facial images of different persons and their corresponding multiple candidate person appearance images; an identity recognition module for identifying the target facial image to obtain an identity recognition result of the person to be identified.

[0014] In a third aspect, the present invention further provides an electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, a pedestrian identification method as described above is implemented.

[0015] In a fourth aspect, the present invention further provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements any of the pedestrian identification methods described above.

[0016] In a fifth aspect, the present invention further provides a computer program product, comprising a computer program, which, when executed by a processor, implements any of the above-described pedestrian identification methods.

[0017] The present invention provides a pedestrian identification method that obtains a person's appearance image and matches it with a target facial image based on a pre-stored personnel archive. The personnel archive includes facial images of different people and their corresponding multiple candidate appearance images. The target facial image is then identified to obtain an identification result for the person to be identified. This method leverages the high accuracy of facial recognition and the passive nature of appearance recognition. By establishing a correlation between the facial image and the appearance image, the identification result for the person to be identified can be obtained by simply obtaining the appearance image of the person to be identified. This process does not require the cooperation of the person to be identified and simultaneously improves the accuracy of person identification. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0019] Figure 1 1 is a flow chart of a pedestrian identification method provided by the present invention;

[0020] Figure 2 It is a structural schematic diagram of the pedestrian identification device provided by the present invention;

[0021] Figure 3 It is a structural schematic diagram of the electronic device provided by the present invention. DETAILED DESCRIPTION

[0022] To make the objectives, technical solutions, and advantages of the present invention more clear, the technical solutions of the present invention will be clearly and completely described below in conjunction with the accompanying drawings. Obviously, the embodiments described are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts shall fall within the scope of protection of the present invention.

[0023] It's important to note that existing person recognition technologies can only achieve high recognition accuracy using high-quality facial images when the person being recognized cooperates. However, the required high-quality facial images are unavailable in many scenarios, such as security and person search, limiting the application of facial recognition alone.

[0024] At present, some research has been proposed on passive identity recognition. These works mainly rely on pedestrian re-identification, that is, using full-body images of pedestrians under surveillance to identify people's identities. However, this recognition method is often difficult to achieve a high accuracy rate due to factors such as image blur, clothing and viewing angle.

[0025] In view of this, the present invention proposes a pedestrian identification method based on the correlation of the person's appearance structure. Specifically, Figure 1 A schematic flow chart of a pedestrian identification method provided by an embodiment of the present invention is shown.

[0026] like Figure 1 As shown, the method includes:

[0027] S110, obtaining an appearance image of the person to be identified;

[0028] S120, matching and obtaining a target facial image corresponding to the person's appearance image based on a pre-stored personnel archive; wherein the personnel archive includes the target facial image of the person to be identified and a plurality of candidate person appearance images;

[0029] S130: Recognize the target face image to obtain an identity recognition result of the person to be recognized.

[0030] It is understandable that in step S110 , it is necessary to obtain an appearance image of the person to be identified.

[0031] The person to be identified is any object whose identity a third party wants to identify, especially an individual who does not actively cooperate in the identification scenario.

[0032] The appearance image of a person refers to any appearance image / appearance outline image / pedestrian image of the person to be identified. For example, it can be a full-body front image, full-body back image, partial front image or partial back image of the person to be identified, and there is no specific limitation here.

[0033] It should be noted that in the scenario of passive identity recognition, it is difficult to directly obtain a frontal and high-resolution facial image of the person to be identified. Problems such as image blur, posture changes, and background changes often lead to the inability to quickly and accurately achieve identity recognition.

[0034] Therefore, in this embodiment, firstly, only the appearance image of the person to be identified is obtained. The image can be directly captured by an image acquisition device or extracted from key frames in a surveillance video, which is not specifically limited here.

[0035] It should be clear that compared with the frontal, high-resolution facial image of the person to be identified, the appearance image of the person to be identified is easier to obtain and the requirements for image quality are slightly lower.

[0036] On the basis of obtaining the appearance image of the person to be identified in step S110, further in step S120, a target face image corresponding to the appearance image of the person is matched based on a pre-stored personnel archive.

[0037] It can be understood that a personnel archive is pre-stored in the system, which includes facial images of all personnel in the target area and appearance images of candidate characters. The correspondence between facial images and appearance images of candidate characters can be one-to-one, one-to-many, or many-to-many, which is not specifically limited here.

[0038] In a specific embodiment, considering that people's clothing and perspectives are relatively variable, there is a one-to-many correspondence between facial images and candidate character appearance images, that is, one facial image corresponds to multiple candidate character appearance images, and these multiple candidate character appearance images are obtained by shooting the corresponding person from different perspectives and different distances.

[0039] After obtaining the appearance image of the person to be identified, the appearance image of the person is matched with multiple appearance images of candidate persons in the personnel archive for similarity. The facial image corresponding to the appearance image of the candidate person with the highest matching degree is the target facial image corresponding to the appearance image of the person.

[0040] Based on the target facial image corresponding to the person's appearance image matched based on the pre-stored personnel archive in step S120, further, in step S130, the target facial image is subjected to identity recognition to obtain an identity recognition result of the person to be recognized.

[0041] It is understandable that, in this step, a pre-trained identity recognition model can be used, with the target face image as the model input, to obtain the output identity recognition information of the person to be identified.

[0042] In addition, since the personnel archive includes facial images of all personnel in the target area and appearance images of candidate persons, all personnel are bound to corresponding identity information. Once the target facial image is determined, the identity information corresponding to the target facial image can be further directly determined, thereby obtaining the identity identification information of the person to be identified.

[0043] The identification result refers to information that can represent the identity of the person to be identified, including but not limited to name, gender, ID number, etc., which is not specifically limited here.

[0044] It should be noted that the pedestrian identification method provided in this embodiment is particularly suitable for passive identification scenarios such as security, traffic statistics, and intelligent person search.

[0045] In this embodiment, a facial image of the person to be identified is obtained and matched to a target facial image based on a pre-stored database of person profiles. The database includes facial images of different people and their corresponding multiple candidate person profiles. The target facial image is then recognized to obtain the identity of the person to be identified. This method leverages the high accuracy of facial recognition and the passive nature of person profile recognition. By establishing a correlation between the facial image and the person profile image, the identity of the person to be identified can be obtained even with the person profile image. This process does not require the cooperation of the person to be identified and improves the accuracy of person identification.

[0046] On the basis of the above embodiment, the process of constructing the personnel archive will be described in detail below.

[0047] Constructing a personnel archive specifically includes: determining the target area and collecting surveillance videos of the target area during a set period of time; extracting facial images corresponding to key frames in the surveillance video based on a face detection algorithm; extracting the appearance images of candidate characters corresponding to key frames in the surveillance video based on a pedestrian detection algorithm; matching the facial images of the same person with the appearance images of candidate characters, and binding the identity information of the person to construct a personnel archive.

[0048] It should be noted that the pedestrian identification method provided in this embodiment is only applicable to persons who have been registered in the personnel archive.

[0049] It is understandable that when building a personnel archive, it is first necessary to determine the target area, such as an office building, a hospital, or an airport. The purpose of determining the target area is to narrow the scope of the personnel to be identified.

[0050] Then, surveillance video of the target area during the set time period is collected, which can be obtained through the surveillance equipment installed in the target area.

[0051] The set time period can be set according to the actual situation of the flow of people, for example, the time period with peak flow of people can be selected as the set time period.

[0052] Next, based on the face detection algorithm, the face image corresponding to the key frame in the surveillance video is extracted, and based on the pedestrian detection algorithm, the appearance image of the candidate person corresponding to the key frame in the surveillance video is extracted.

[0053] Among them, common face detection algorithms include Haar, Hog, CNN (Convolutional Neural Network), SSD (Single Shot MultiBox Detector), and MTCNN (Multi-task convolutional neural network).

[0054] Pedestrian detection algorithms include feature-based methods, deep learning-based methods, and target tracking-based methods.

[0055] Among them, feature-based methods usually extract features from images, such as edges, colors, textures, etc., and use these features to detect pedestrians; deep learning-based methods usually automatically learn features in images through convolutional neural networks or recurrent neural networks, and have good generalization capabilities; target tracking-based methods usually use target tracking technology to detect pedestrians. Target tracking can estimate the position and posture of pedestrians through continuous image frames.

[0056] It should be noted that since face detection algorithms and pedestrian detection algorithms are already relatively mature technologies, they will not be elaborated here.

[0057] A key frame refers to an image captured from a surveillance video that contains important content (such as pedestrians and faces).

[0058] On the basis of extracting the facial image and the appearance image of the candidate person, the facial image and the appearance image of the candidate person of the same person are further matched and bound with the identity information of the person to obtain the complete image information of each person.

[0059] The facial image of the same person is matched with the appearance image of the candidate person. Specifically, when a facial image of a certain person in the personnel archive is completely contained in the appearance image of a certain candidate person, a matching relationship between the facial image and the appearance image of the candidate person can be established.

[0060] After matching the facial image of the same person with the appearance image of the candidate character, denoising is required, that is, removing possible erroneous matching relationships.

[0061] Specifically, when a certain facial image contains candidate person appearance images corresponding to multiple persons (for example, multiple pedestrians), the matching relationship between the facial image and these candidate person appearance images is cancelled.

[0062] And / or, when a candidate character appearance image contains facial images corresponding to multiple persons, the matching relationship between these facial images and the candidate character appearance image is cancelled.

[0063] Based on the above, a personnel archive of the target area in the set time period can be constructed.

[0064] It should be noted that in addition to extracting facial images and candidate person appearance images from surveillance videos, it is also possible to collect images at the entrance of the target area. For every person who enters the target area for the first time, a camera or other image acquisition device will collect their high-quality facial images and candidate person appearance images from multiple angles and different distances.

[0065] In this embodiment, a target area is determined and surveillance video of the target area is collected during a set period. Then, based on a face detection algorithm, facial images corresponding to key frames in the surveillance video are extracted. Furthermore, based on a pedestrian detection algorithm, the appearance images of candidate individuals corresponding to key frames in the surveillance video are extracted. The facial images of the same individual are matched with the appearance images of the candidate individuals and bound to the individual's identity information to construct a personnel archive. Based on the personnel archive, the target facial images corresponding to the appearance images of the individuals are matched, thereby identifying the target facial images and obtaining the identity recognition results of the individual to be identified. This method fully utilizes the high accuracy of face recognition and the passive nature of appearance recognition. By establishing an association between facial images and appearance images of individuals, the identity recognition results of the individual to be identified can be obtained by obtaining the appearance images of the individual to be identified. This process does not require the cooperation of the individual to be identified and simultaneously improves the accuracy of individual identification.

[0066] On the basis of the above embodiment, the process of matching and obtaining a target face image corresponding to a person's appearance image will be described in detail below.

[0067] Based on a pre-stored personnel archive, a target facial image corresponding to a person's appearance image is matched, including: extracting a first person appearance feature from the person's appearance image; extracting a second person appearance feature corresponding to each candidate person's appearance image in the personnel archive; calculating the cosine similarity between the first person appearance feature and each second person appearance feature; determining the maximum cosine similarity among all cosine similarities, and obtaining a facial image corresponding to the candidate person's appearance image corresponding to the maximum cosine similarity to obtain a target facial image.

[0068] It is understandable that the appearance image of the person to be identified and multiple appearance images of candidate persons in the personnel archive are known, and each appearance image of the candidate person has a corresponding face image.

[0069] In this case, features are extracted from the appearance image of the person to be identified to obtain first appearance features of the person. At the same time, features are extracted from the appearance image of each candidate person in the personnel archive to obtain multiple second appearance features of the person.

[0070] Furthermore, the cosine similarity between the first person appearance feature and each second person appearance feature is calculated to obtain a plurality of cosine similarities.

[0071] Next, when the cosine similarity is higher than the set threshold, a facial image corresponding to the candidate person's appearance image corresponding to the cosine similarity is obtained, and the facial image is the target facial image.

[0072] Normally, there is only one cosine similarity above the set threshold.

[0073] If there are multiple cosine similarities that are higher than the set threshold, the maximum cosine similarity among the multiple cosine similarities is selected, and the corresponding candidate person appearance image and the facial image corresponding to the candidate person appearance image are obtained based on the maximum cosine similarity. The facial image is the target facial image.

[0074] The threshold value may be set according to actual needs. For example, in a specific embodiment, the threshold value is set to 95%.

[0075] Then, identity recognition is performed on the target face image to obtain the identity recognition result of the person to be identified.

[0076] In this embodiment, by extracting a first person appearance feature from a person's appearance image, and extracting a second person appearance feature corresponding to each candidate person appearance image in the person archive, the cosine similarity between the first person appearance feature and each second person appearance feature is calculated, and the maximum cosine similarity among all cosine similarities is determined. The facial image corresponding to the candidate person appearance image corresponding to the maximum cosine similarity is then obtained to obtain a target facial image. Thus, the target facial image is recognized to obtain an identification result of the person to be identified. This method fully utilizes the high accuracy of facial recognition and the passive nature of person appearance recognition. By establishing an association between the facial image and the person appearance image, the identification result of the person to be identified can be obtained by obtaining the person appearance image of the person to be identified. This process does not require the cooperation of the person to be identified, while improving the accuracy of person identification.

[0077] In one embodiment, for example, the target area is an airport lounge. First, each passenger's facial image and candidate person appearance image are captured as they enter the lounge, and their identity information is associated with the image. Simultaneously, facial images and candidate person appearance images from several consecutive key frames in the surveillance video are extracted and added to the personnel profile database, resulting in a constructed personnel profile.

[0078] When performing personnel identity recognition, the acquired appearance image of the person to be identified is matched with the appearance images of candidate persons in the personnel archive, and the facial image corresponding to the appearance image of the candidate person whose cosine similarity is greater than the set threshold is taken as the target facial image of the person to be identified.

[0079] Perform identity recognition on the target face image to obtain the identity recognition result.

[0080] Furthermore, subsequent operations can be performed based on the identity recognition results, such as reminding the passenger when he is about to check in.

[0081] After that, the passengers who have boarded the plane can be promptly de-stocked to avoid misidentification of the facial images of passengers who have left.

[0082] It should be noted that the personnel archive can be updated in real time according to the actual personnel situation of the airport to achieve continuous real-time application.

[0083] The pedestrian identification method provided in the embodiment of the present invention aims to achieve high-reliability personnel identification in scenarios where the person being identified is unable to actively cooperate, and has broad application prospects in fields such as intelligent security.

[0084] The pedestrian identification method proposed in the embodiment of the present invention utilizes the high accuracy of face recognition and the passive characteristics of pedestrian recognition. Compared with using only pedestrian body images for recognition, the accuracy will be greatly improved.

[0085] The embodiment of the present invention uses the appearance image of a person to associate with the face image. Subsequently, with the help of the currently very complete face recognition system and large database, the identity of the recognized person can be directly retrieved, which can bring greater convenience to smart security, smart search for people and other aspects.

[0086] Figure 2 A schematic structural diagram of a pedestrian identification device provided by an embodiment of the present invention is shown.

[0087] like Figure 2 As shown, the device includes:

[0088] An image acquisition module 210 is used to acquire an apparent image of a person to be identified;

[0089] An image matching module 220 is configured to match a target facial image corresponding to the person's appearance image based on a pre-stored personnel archive, wherein the personnel archive includes facial images of different persons and their corresponding multiple candidate person appearance images;

[0090] The identity recognition module 230 is used to recognize the target face image and obtain the identity recognition result of the person to be recognized.

[0091] In this embodiment, the image acquisition module 210 acquires the appearance image of the person to be identified. The image matching module 220 matches the appearance image of the person to be identified with a target facial image based on a pre-stored personnel archive. The personnel archive includes facial images of different persons and their corresponding appearance images of multiple candidate persons. The identity recognition module 230 then recognizes the target facial image to obtain an identification result for the person to be identified. This device fully utilizes the high accuracy of facial recognition and the passive nature of appearance recognition. By establishing a correlation between the facial image and the appearance image of the person to be identified, the identification result of the person to be identified can be obtained by acquiring the appearance image of the person to be identified. This process does not require the cooperation of the person to be identified, while also improving the accuracy of person identification.

[0092] It should be noted that the pedestrian identification device provided in this embodiment and the pedestrian identification method described above can be referred to in correspondence with each other, and will not be described in detail here.

[0093] Figure 3 An example of a physical structure diagram of an electronic device is shown below. Figure 3As shown, the electronic device may include: a processor 310, a communication interface 320, a memory 330, and a communication bus 340, wherein the processor 310, the communication interface 320, and the memory 330 communicate with each other via the communication bus 340. The processor 310 may call the logic instructions in the memory 330 to execute a pedestrian identification method, which includes: obtaining a person appearance image of a person to be identified; matching a target face image corresponding to the person appearance image based on a pre-stored person profile library; wherein the person profile library includes face images of different persons and their corresponding multiple candidate person appearance images; and identifying the target face image to obtain an identification result of the person to be identified.

[0094] In addition, the logic instructions in the above-mentioned memory 330 can be implemented in the form of a software functional unit and can be stored in a computer-readable storage medium when sold or used as an independent product. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions for enabling a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: various media that can store program codes, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk.

[0095] On the other hand, the present invention also provides a computer program product, which includes a computer program, which can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer can execute the pedestrian identification method provided by the above methods, which includes: obtaining a person appearance image of the person to be identified; based on a pre-stored person archive, matching to obtain a target face image corresponding to the person appearance image; wherein the person archive includes face images of different persons and their corresponding multiple candidate person appearance images; identifying the target face image to obtain the identity recognition result of the person to be identified.

[0096] On the other hand, the present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon. When the computer program is executed by a processor, it is implemented to execute the pedestrian identification method provided by the above-mentioned methods, the method comprising: obtaining a person appearance image of the person to be identified; based on a pre-stored personnel archive, matching and obtaining a target facial image corresponding to the person appearance image; wherein the personnel archive includes facial images of different persons and their corresponding multiple candidate person appearance images; identifying the target facial image to obtain the identity identification result of the person to be identified.

[0097] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, i.e., they may be located in one location or distributed across multiple network units. Some or all of the modules may be selected based on actual needs to achieve the objectives of the present embodiment. Persons of ordinary skill in the art will be able to understand and implement the present invention without inventive effort.

[0098] Through the description of the above embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus a necessary general hardware platform, or of course, by hardware. Based on this understanding, the essence of the above technical solution or the part that contributes to the existing technology can be embodied in the form of a software product. The computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a magnetic disk, an optical disk, etc., and includes a number of instructions for enabling a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or certain parts of the embodiments.

[0099] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the various embodiments of the present invention.

Claims

1. A pedestrian identification method, characterized in that: include: Obtaining an apparent image of a person to be identified; Based on a pre-stored personnel archive, a target facial image corresponding to the person's appearance image is matched; wherein the personnel archive includes facial images of different persons and their corresponding multiple candidate person appearance images; The target face image is recognized to obtain an identity recognition result of the person to be recognized.

2. The pedestrian identification method according to claim 1, characterized in that: Constructing the personnel archive library specifically includes: Determine the target area and collect surveillance video of the target area during a set period of time; Extracting facial images corresponding to key frames in the surveillance video based on a face detection algorithm; Extracting the candidate person's appearance image corresponding to the key frame in the surveillance video based on the pedestrian detection algorithm; The facial image of the same person and the appearance image of the candidate person are matched and bound to the identity information of the person to construct the person archive.

3. The pedestrian identification method according to claim 1, characterized in that: The matching based on the pre-stored personnel archive to obtain the target face image corresponding to the person's appearance image includes: extracting a first person appearance feature of the person appearance image; Extracting a second person appearance feature corresponding to each candidate person appearance image in the personnel archive; Calculating the cosine similarity between the first person appearance feature and each of the second person appearance features; When the cosine similarity is higher than a set threshold, a facial image corresponding to the candidate person's appearance image corresponding to the cosine similarity is obtained to obtain the target facial image.

4. The pedestrian identification method according to claim 2, characterized in that: The matching of the facial image of the same person and the appearance image of the candidate person includes: In the case that the face image is completely contained in the candidate character appearance image, a matching relationship is established between the face image and the candidate character appearance image.

5. The pedestrian identification method according to claim 4, characterized in that: The matching of the facial image of the same person and the appearance image of the candidate person further includes: In a case where the facial image contains candidate person appearance images corresponding to multiple persons, canceling the matching relationship between the facial image and the candidate person appearance images; and / or, In a case where the candidate character appearance image includes facial images corresponding to multiple persons, the matching relationship between the facial images and the candidate character appearance image is cancelled.

6. The pedestrian identification method according to any one of claims 1 to 5, characterized in that: The person appearance image is a full-body front image, a full-body back image, a partial front image or a partial back image of the person to be identified.

7. A pedestrian identification device, characterized in that: include: An image acquisition module is used to acquire an image of the person to be identified; An image matching module is configured to match a target facial image corresponding to the person's appearance image based on a pre-stored personnel archive, wherein the personnel archive includes facial images of different persons and their corresponding multiple candidate person appearance images; The identity recognition module is used to recognize the target face image and obtain the identity recognition result of the person to be recognized.

8. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the program, the pedestrian identification method according to any one of claims 1 to 6 is implemented.

9. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the pedestrian identification method according to any one of claims 1 to 6 is implemented.

10. A computer program product comprising a computer program, characterized in that When the computer program is executed by a processor, the pedestrian identification method according to any one of claims 1 to 6 is implemented.

Images