WO2021082118A1

WO2021082118A1 - Person re-identification method and apparatus, and terminal and storage medium

Info

Publication number: WO2021082118A1
Application number: PCT/CN2019/119860
Authority: WO
Inventors: 李国法; 黄莉莎; 徐刚; 谢恒�; 赖伟鉴; 陈耀昱
Original assignee: 深圳大学
Priority date: 2019-11-01
Filing date: 2019-11-21
Publication date: 2021-05-06
Also published as: CN111027378B; CN111027378A

Abstract

The present application is applicable to the technical field of computers. Provided is a person re-identification method, comprising: acquiring a target image frame sequence from a pre-collected monitoring video stream; identifying, from image information, feature information of a person to be identified, and determining all pieces of tag information corresponding to the feature information; respectively performing person re-identification on the target image frame sequence and all the pieces of tag information by means of a pre-trained person re-identification model, and determining, from all the pieces of tag information, target tag information of the person to be identified; and determining, on the basis of the target tag information, a re-identification result of the person to be identified. By means of identifying, from image information including a person to be identified, feature information of a person to be identified, determining all pieces of tag information corresponding to the feature information, and respectively performing person re-identification on the image information including the person to be identified and all the pieces of tag information by means of a pre-trained person re-identification model, the accuracy of person re-identification is improved.

Description

Method, device, terminal and storage medium for pedestrian re-identification

This application claims the priority of the Chinese patent application filed at the Chinese Patent Office on November 1, 2019, with the application number 201911060337.0, and the invention title "Methods, devices, terminals and storage media for pedestrian re-identification", and its entire contents Incorporated in this application by reference.

Technical field

This application belongs to the field of computer technology, and in particular relates to a method, device, terminal, and storage medium for pedestrian re-identification.

Background technique

Pedestrian re-identification (Person re-identification), also known as pedestrian re-identification, is a technology that uses computer vision technology to determine whether there is a specific pedestrian in an image or video sequence. At present, in the process of pedestrian re-identification, network recognition models are often used. However, due to the differences between different camera equipment, pedestrians have both rigid and flexible characteristics, and the appearance is easily affected by wearing, scale, occlusion, posture and viewing angle, making the process of pedestrian re-recognition more difficult than the common face recognition process How to improve the accuracy of pedestrian re-identification is a technical problem that needs to be solved urgently.

Summary of the invention

In view of this, the embodiments of the present application provide a pedestrian re-identification method, device, terminal, and storage medium to improve the accuracy of pedestrian re-identification.

The first aspect of the embodiments of the present application provides a pedestrian re-identification method, including:

Acquiring a target image frame sequence from a pre-collected surveillance video stream, where the target image frame in the target image frame sequence contains image information of a pedestrian to be identified;

Identifying characteristic information of the pedestrian to be identified from the image information, and determining all tag information corresponding to the characteristic information;

Using a pre-trained pedestrian re-recognition model to perform pedestrian re-recognition on the target image frame sequence and all the tag information, and determine the target tag information of the pedestrian to be identified from all the tag information;

The re-identification result of the pedestrian to be identified is determined based on the target tag information.

A second aspect of the embodiments of the present application provides a pedestrian re-identification device, including:

An acquiring module, configured to acquire a target image frame sequence from a pre-collected surveillance video stream, and the target image frame in the target image frame sequence contains image information of a pedestrian to be identified;

The first determining module is configured to identify the characteristic information of the pedestrian to be identified from the image information, and determine all tag information corresponding to the characteristic information;

The re-recognition module is used to perform pedestrian re-recognition on the target image frame sequence and all the tag information using the pre-trained pedestrian re-recognition model, and determine the target tag of the pedestrian to be recognized from all the tag information information;

The second determination module is configured to determine the re-identification result of the pedestrian to be identified based on the target tag information.

A third aspect of the embodiments of the present application provides a terminal, including a memory, a processor, and a computer program stored in the memory and running on the processor, wherein the processor executes the computer program When realizing the steps of the pedestrian re-identification method described in the first aspect of the above embodiment.

The fourth aspect of the embodiments of the present application provides a computer-readable storage medium, the computer-readable storage medium stores a computer program, and is characterized in that, when the computer program is executed by a processor, the implementation is as described in the first aspect of the above embodiment. Steps of the pedestrian re-identification method.

The details of one or more embodiments of the present application are set forth in the following drawings and description. Other features, purposes and advantages of this application will become apparent from the description, drawings and claims.

Description of the drawings

In order to more clearly describe the technical solutions in the embodiments of the present application, the following will briefly introduce the drawings that need to be used in the description of the embodiments or the prior art. Obviously, the drawings in the following description are only of the present application. For some embodiments, those of ordinary skill in the art can obtain other drawings based on these drawings without creative labor.

Fig. 1 is an implementation flowchart of the pedestrian re-identification method provided by the first embodiment of the present application;

Figure 2 is a flow chart of the specific implementation of S102 in Figure 1;

Fig. 3 is an implementation flowchart of the pedestrian re-identification method provided by the second embodiment of the present application;

Figure 4 is a flowchart of the specific implementation of S304 in Figure 3;

Figure 5 is a specific implementation flow chart of S305 in Figure 3;

FIG. 6 is a schematic structural diagram of a pedestrian re-identification device provided by an embodiment of the present application;

FIG. 7 is a schematic structural diagram of a terminal provided by an embodiment of the present application.

Detailed ways

In the following description, for the purpose of illustration rather than limitation, specific details such as a specific system structure and technology are proposed for a thorough understanding of the embodiments of the present application. However, it should be clear to those skilled in the art that the present application can also be implemented in other embodiments without these specific details. In other cases, detailed descriptions of well-known systems, devices, circuits, and methods are omitted to avoid unnecessary details from obstructing the description of this application.

It should be noted that with the rapid development of machine (deep) learning and the increasing popularity of video surveillance equipment, pedestrian re-identification has been paid more and more attention in the fields of intelligent security and intelligent surveillance. Most existing pedestrian re-identification methods use machine learning models to identify whether the pedestrians in the image frame are the same person, and regard the pedestrian re-identification problem as a multi-classification problem. The loss function of common machine learning models used for classification is usually a cross-entropy loss function. Since the cross-entropy loss function only calculates the loss between the training sample and the correct category to ensure the correctness of the classification, and does not consider the loss information of the misjudgment, there is a certain error in the classification result. Therefore, how to integrate the loss of the wrong category into the loss function, reduce the probability of misjudgment, and improve the performance of the pedestrian re-identification network model is an urgent problem to be solved.

The present invention provides a pedestrian re-identification method, which performs pedestrian re-identification based on a pedestrian re-identification network model of a novel loss function, and improves the cross-entropy loss function by increasing the loss of the wrong category, thereby enabling the machine learning model to obtain better classification performance .

In order to illustrate the technical solution described in the present application, specific embodiments are used for description below. As shown in Figure 1, it is a flow chart of the implementation of the pedestrian re-identification method provided by the first embodiment of the present application. This embodiment may be implemented by hardware or software of a pedestrian re-identification device. The pedestrian re-identification device may be terminal. The details are as follows:

S101. Obtain a target image frame sequence from a pre-collected surveillance video stream, where the target image frame in the target image frame sequence contains image information of a pedestrian to be identified.

The pre-collected surveillance video stream is collected by a pre-determined surveillance device, such as a video stream collected by a surveillance device in a campus, the surveillance video stream includes consecutive image frames in a time sequence, and the video streams collected in different time periods include The image frame correspondingly contains different target information. In this embodiment, a target image frame sequence is acquired from a pre-collected surveillance video stream, and the target image frame in the target image frame sequence contains the image information of the pedestrian to be identified, and the image information of the pedestrian to be identified includes the image information of the pedestrian to be identified. Recognize pedestrian's facial information, clothing information, body information, etc.

S102: Identify feature information of the pedestrian to be identified from the image information, and determine all tag information corresponding to the feature information.

The feature information of the pedestrian to be identified includes facial features such as skin condition, facial expression, facial features, appearance features such as clothes color, clothes texture, handbag, backpack, hat, etc., the image area occupied, and the relative position in the image. Location features such as location. In this embodiment, the characteristic information has corresponding label information, and the label information is used to identify the characteristic information. For example, the characteristic information is a skin condition, and the corresponding label information is smooth or not smooth.

In an optional implementation manner, as shown in FIG. 2, it is a flowchart of the specific implementation of S102 in FIG. 1. It can be seen from Figure 2 that S102 includes:

S1021: Perform feature recognition on the pedestrian to be identified by using the feature information recognition model completed in advance to obtain the feature information of the pedestrian to be identified.

The pre-trained feature information recognition model may be a machine learning model with a recognition function, such as a neural network model. The input of the neural network model is the pedestrian to be recognized, and the output is the feature information corresponding to the pedestrian to be recognized.

S1022: Calculate the probability value of the feature information belonging to each type of preset label information.

Understandably, the pedestrian to be identified usually has multiple different characteristic information, such as skin condition, clothes color, etc., and different characteristic information corresponds to multiple preset label information, for example, the preset label information corresponding to the clothes color includes red and black. , Yellow, green, etc., and when the feature information is not obvious, multiple preset label information may appear in the recognition result. For example, when the color of clothes is black, the recognition result may correspond to two preset labels of black and gray. At this time, it is necessary to further calculate the probability value of the feature information belonging to each type of preset label information.

Specifically, the probability value of the feature information belonging to each type of preset label information can be calculated by using a preset probability normalization formula; the preset probability normalization formula is:

Wherein, p _i represents the probability value of the feature information belonging to the i-th type of preset label information, and K represents the total number of types of preset label information,

It indicates that the feature information belongs to the log probability value of the i-th type of preset label information.

S1023: If the probability value of the feature information belonging to the first type of preset label information is greater than the probability value of belonging to the second type of preset tag information, and the probability value of the feature information belonging to the first type of preset label information is greater than A preset probability threshold, it is determined that the first type of preset label information is the label information corresponding to the feature information, and the second type of preset label information is other than the first type of preset label information Any type of preset label information.

In this embodiment, the preset tag information corresponding to the feature information is determined by calculating the probability value that the feature information belongs to the preset tag information. Understandably, the method of calculating the probability value is not limited to using the aforementioned preset probability normalization formula, which is not specifically limited here.

S103, using a pre-trained pedestrian re-recognition model to perform pedestrian re-identification on the target image frame sequence and all the tag information, and determine the target tag information of the pedestrian to be identified from all the tag information.

The pre-trained pedestrian re-recognition model may be a machine learning model with a recognition function, and the input of the pedestrian re-recognition model is an image frame sequence containing the pedestrian to be recognized and all tags corresponding to the characteristic information of the pedestrian to be recognized Information, output as target tag information of the pedestrian to be identified. Wherein, the pedestrian to be identified belongs to tag information with the highest probability of each type of preset tag information.

S104: Determine a re-identification result of the pedestrian to be identified based on the target tag information.

Specifically, when the target tag information matches the preset tag information of the pedestrian to be identified, it is determined that the pedestrian to be identified is a predetermined specific pedestrian. If the tag information for identifying the pedestrian does not match, it is determined that the pedestrian to be identified is not a predetermined specific pedestrian.

From the above analysis, it can be seen that the pedestrian recovery method provided by the embodiment of the present application includes: acquiring a target image frame sequence from a pre-collected surveillance video stream, and the target image frame in the target image frame sequence contains the pedestrian to be identified Identify the feature information of the pedestrian to be identified from the image information, and determine all the tag information corresponding to the feature information; use the pre-trained pedestrian re-recognition model to separately perform the target image frame sequence Perform pedestrian re-identification with all the tag information, determine the target tag information of the pedestrian to be identified from all the tag information; determine the re-identification result of the pedestrian to be identified based on the target tag information. Compared with the prior art, by identifying the characteristic information of the pedestrian to be identified from the image information containing the pedestrian to be identified, all the tag information corresponding to the characteristic information is determined; Pedestrian image information and all tag information are re-identified to improve the accuracy of pedestrian re-identification.

As shown in FIG. 3, it is a flowchart of the implementation of the pedestrian re-identification method provided by the second embodiment of the present application. It can be seen from Fig. 3 that compared with the embodiment shown in Fig. 1, the specific implementation process of S301~S302 and S308~S309 is the same as the specific implementation process of S101~S104 in this implementation. The difference is that S303 is also included before S308. ~ S307, where S307 and S308 are executed in parallel, and one of them can be executed. The specific implementation process of S303~S307 is detailed as follows:

S303: Collect a first preset number of training samples, each of which includes an image of a pedestrian to be identified and all preset label information corresponding to the pedestrian to be identified.

S304: Use the training sample to train a pre-established machine learning model for training, to obtain a trained machine learning model.

As shown in FIG. 4, it is a flowchart of the specific implementation of S304 in FIG. 3. It can be seen from Figure 4 that S304 includes:

S3041. Use the pre-established machine learning model to re-identify all preset label information corresponding to each pedestrian to be identified, obtain the probability that each pedestrian to be identified belongs to each type of preset label information, and determine each The preset tag information with the highest probability corresponding to the pedestrian to be identified.

The machine learning model may be a deep learning model such as a neural network model, a logical classification model, and a random forest model. It is understandable that all the preset label information corresponding to each pedestrian to be identified are usually not completely the same. Performing preset label recognition through the machine learning model can quickly and accurately obtain that each pedestrian to be identified belongs to each category. The probability of the preset label information.

S3042, respectively using the preset label information with the highest probability corresponding to each pedestrian to be identified as a constraint condition for training the machine learning model, and iterating the preset parameters of the machine learning model.

Understandably, the preset label property with the highest probability corresponding to each pedestrian to be identified is used as a constraint condition for training the machine learning model, and by minimizing the loss function corresponding to the machine learning model, iteratively The preset parameters of the machine learning model are used to improve the accuracy of the machine learning model's recognition of error labels.

S3043: If the rate of change of the loss function value corresponding to the machine learning model tends to be stable, it is determined that the training of the machine learning model is completed, and the pedestrian re-identification model is obtained.

S305: Perform a model accuracy test on the machine learning model after training.

As shown in FIG. 5, it is a specific implementation flowchart of S305 in FIG. 3. It can be seen from Figure 5 that S305 includes:

S3051: Input a second preset number of test samples into the machine learning model after training for analysis, and determine the rate of change of the loss function of the machine learning model after training.

S3052: If the rate of change is less than or equal to a preset rate of change threshold, it is determined that the test of the machine learning model after training passes.

S3053: If the rate of change is greater than a preset rate of change threshold, it is determined that the test of the machine learning model after the training fails.

S306: If the accuracy test of the machine learning model after the training is passed, it is determined that the machine learning model after the training is the pedestrian re-identification model.

S307: If the accuracy test of the machine learning model after the training fails, increase the number of training samples, and return to perform training using the training sample to train the pre-established machine learning model to obtain all The pedestrian re-identification model.

The loss function of the pre-trained pedestrian re-identification model is:

among them,

Where N represents the total number of training samples, K represents the total number of categories of preset label information, p _j represents the probability value of the current sample belonging to the j-th type of preset label information, y _i is the true label information corresponding to the current sample, q _{i, j} are the distribution ratios of p _j _{, N sc} represents the number of similar label information belonging to the current sample, and ε is a coefficient that balances the real label information and the similar label information,

Indicates that the feature information corresponding to the current sample belongs to the log probability value of the j-th type of preset label information,

Indicates that the feature information corresponding to the current sample belongs to the log probability value of the k-th type of preset label information,

Represents the first n tags whose output probability is greater than the preset probability threshold.

Fig. 6 is a schematic structural diagram of a pedestrian re-identification device provided by an embodiment of the present application. It can be seen from FIG. 6 that the pedestrian re-identification device 6 provided in this embodiment includes: an acquisition module 601, a first determination module 602, a re-identification module 603, and a second determination module 604. among them,

The obtaining module 601 is configured to obtain a target image frame sequence from a pre-collected surveillance video stream, and the target image frame in the target image frame sequence contains image information of a pedestrian to be identified;

The first determining module 602 is configured to identify the characteristic information of the pedestrian to be identified from the image information, and determine all tag information corresponding to the characteristic information;

The re-recognition module 603 is configured to use the pre-trained pedestrian re-recognition model to perform pedestrian re-recognition on the target image frame sequence and all the tag information, and determine the target of the pedestrian to be recognized from all the tag information Label Information;

The second determining module 604 is configured to determine the re-identification result of the pedestrian to be identified based on the target tag information.

In an optional implementation manner, the first determining module 602 includes:

A recognition unit, configured to use a pre-trained feature information recognition model to perform feature recognition on the pedestrian to be identified to obtain feature information of the pedestrian to be identified;

A calculation unit, configured to calculate the probability value of the feature information belonging to each type of preset label information;

The first determining unit is configured to: if the probability value of the feature information belonging to the first type of preset label information is greater than the probability value of belonging to the second type of preset label information, and the feature information belongs to the first type of preset If the probability value of the tag information is greater than the preset probability threshold, it is determined that the first type of preset tag information is the tag information corresponding to the feature information, and the second type of preset tag information is except for the first type of preset tag information. Set any type of preset label information except label information.

In an optional implementation manner, the calculation unit includes:

A preset probability normalization formula is used to calculate the probability value of the feature information belonging to each type of preset label information; the preset probability normalization formula is:

Indicates the log probability value of the feature information belonging to the i-th type of preset label information.

In an optional implementation manner, it further includes:

An acquisition module, configured to collect a first preset number of training samples, each of the training samples includes an image of the pedestrian to be identified and all preset label information corresponding to the pedestrian to be identified;

The training module is used for training a pre-established machine learning model by using the training samples to obtain a machine learning model after training;

A test module, which is used to perform a model accuracy test on the machine learning model after training;

The first determination module is configured to determine that the machine learning model after the training is the pedestrian re-identification model if the accuracy test of the machine learning model after the training is passed;

The second determination module is used to increase the number of training samples if the accuracy test of the machine learning model after the training fails, and then return to execute the machine learning pre-established for training with the training samples The model is trained to obtain the pedestrian re-identification model.

In an optional implementation manner, the training module includes:

The re-identification unit is configured to use the pre-established machine learning model to re-identify all the preset label information corresponding to each pedestrian to be identified, and obtain the probability that each pedestrian to be identified belongs to each type of preset label information, And determine the preset label information with the highest probability corresponding to each of the pedestrians to be identified;

An iterative unit, configured to use the preset label information with the highest probability corresponding to each pedestrian to be identified as a constraint condition for training the machine learning model, and iterate the preset parameters of the machine learning model;

The second determining unit is configured to determine that the training of the machine learning model is completed if the rate of change of the loss function value corresponding to the machine learning model becomes stable, and the pedestrian re-identification model is obtained.

In an optional implementation manner, the loss function of the pre-trained pedestrian re-identification model is:

among them,

FIG. 7 is a schematic structural diagram of a terminal provided by an embodiment of the present application. As shown in FIG. 7, the terminal 7 of this embodiment includes a processor 70, a memory 71, and a computer program 72 stored in the memory 71 and running on the processor 70, such as a pedestrian re-identification program. When the processor 70 executes the computer program 72, the steps in the above-mentioned various pedestrian re-identification method embodiments are implemented, for example, steps 101 to 104 shown in FIG. 1.

Exemplarily, the computer program 72 may be divided into one or more modules/units, and the one or more modules/units are stored in the memory 71 and executed by the processor 70 to complete the application. The one or more modules/units may be a series of computer program instruction segments capable of completing specific functions, and the instruction segments are used to describe the execution process of the computer program 72 in the terminal 7. For example, the computer program 72 can be divided into an acquisition module, a first determination module, a re-identification module, and a second determination module (a module in a virtual device). The specific functions of each module are as follows:

Those skilled in the art can clearly understand that, for the convenience and conciseness of description, only the division of the above functional units and modules is used as an example. In practical applications, the above functions can be allocated to different functional units and modules as needed. Module completion, that is, the internal structure of the device is divided into different functional units or modules to complete all or part of the functions described above. The functional units and modules in the embodiments can be integrated into one processing unit, or each unit can exist alone physically, or two or more units can be integrated into one unit. The above-mentioned integrated units can be hardware-based Formal realization can also be realized in the form of software functional units. In addition, the specific names of the functional units and modules are only used to facilitate distinguishing each other, and are not used to limit the protection scope of the present application. For the specific working process of the units and modules in the foregoing system, reference may be made to the corresponding process in the foregoing method embodiment, which will not be repeated here.

In the above-mentioned embodiments, the description of each embodiment has its own emphasis. For parts that are not described in detail or recorded in an embodiment, reference may be made to related descriptions of other embodiments.

A person of ordinary skill in the art may realize that the units and algorithm steps of the examples described in combination with the embodiments disclosed herein can be implemented by electronic hardware or a combination of computer software and electronic hardware. Whether these functions are performed by hardware or software depends on the specific application and design constraint conditions of the technical solution. Professionals and technicians can use different methods for each specific application to implement the described functions, but such implementation should not be considered beyond the scope of this application.

In the embodiments provided in this application, it should be understood that the disclosed device/terminal device and method may be implemented in other ways. For example, the device/terminal device embodiments described above are only illustrative. For example, the division of the modules or units is only a logical function division, and there may be other divisions in actual implementation, such as multiple units. Or components can be combined or integrated into another system, or some features can be omitted or not implemented. In addition, the displayed or discussed mutual coupling or direct coupling or communication connection may be indirect coupling or communication connection through some interfaces, devices or units, and may be in electrical, mechanical or other forms.

The units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple communication units. Some or all of the units may be selected according to actual needs to achieve the objectives of the solutions of the embodiments.

In addition, the functional units in the various embodiments of the present application may be integrated into one processing unit, or each unit may exist alone physically, or two or more units may be integrated into one unit. The above-mentioned integrated unit can be implemented in the form of hardware or software functional unit.

If the integrated module/unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer readable storage medium. Based on this understanding, the present application implements all or part of the processes in the above-mentioned embodiments and methods, and can also be completed by instructing relevant hardware through a computer program. The computer program can be stored in a computer-readable storage medium. When the program is executed by the processor, it can implement the steps of the foregoing method embodiments. . Wherein, the computer program includes computer program code, and the computer program code may be in the form of source code, object code, executable file, or some intermediate forms. The computer-readable medium may include: any entity or device capable of carrying the computer program code, recording medium, U disk, mobile hard disk, magnetic disk, optical disk, computer memory, read-only memory (ROM, Read-Only Memory) , Random Access Memory (RAM, Random Access Memory), electrical carrier signal, telecommunications signal, and software distribution media, etc. It should be noted that the content contained in the computer-readable medium can be appropriately added or deleted according to the requirements of the legislation and patent practice in the jurisdiction. For example, in some jurisdictions, according to the legislation and patent practice, the computer-readable medium Does not include electrical carrier signals and telecommunication signals.

The above-mentioned embodiments are only used to illustrate the technical solutions of the present application, not to limit them; although the present application has been described in detail with reference to the foregoing embodiments, a person of ordinary skill in the art should understand that it can still implement the foregoing The technical solutions recorded in the examples are modified, or some of the technical features are equivalently replaced; these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the application, and should be included in Within the scope of protection of this application.

Claims

A method for pedestrian re-identification, which is characterized in that it includes:

Acquiring a target image frame sequence from a pre-collected surveillance video stream, where the target image frame in the target image frame sequence contains image information of a pedestrian to be identified;

Identifying characteristic information of the pedestrian to be identified from the image information, and determining all tag information corresponding to the characteristic information;

Using a pre-trained pedestrian re-recognition model to perform pedestrian re-recognition on the target image frame sequence and all the tag information, and determine the target tag information of the pedestrian to be identified from all the tag information;

The re-identification result of the pedestrian to be identified is determined based on the target tag information.
The method for pedestrian re-identification according to claim 1, wherein the identifying characteristic information of the pedestrian to be identified from the image information and determining the label information corresponding to the characteristic information comprises:

Using the pre-trained feature information recognition model to perform feature recognition on the pedestrian to be identified to obtain the feature information of the pedestrian to be identified;

Calculating the probability value of the feature information belonging to each type of preset label information;

If the probability value of the feature information belonging to the first type of preset tag information is greater than the probability value of belonging to the second type of preset tag information, and the probability value of the feature information belonging to the first type of preset tag information is greater than the preset The probability threshold of the first type is determined to be the tag information corresponding to the feature information, and the second type of preset tag information is any one other than the first type of preset tag information. Class preset label information.
The method for pedestrian re-identification according to claim 2, wherein said calculating the probability value of said characteristic information belonging to each type of preset label information comprises:

A preset probability normalization formula is used to calculate the probability value of the feature information belonging to each type of preset label information; the preset probability normalization formula is:

Wherein, p i represents the probability value of the feature information belonging to the i-th type of preset label information, and K represents the total number of types of preset label information,
Indicates the log probability value of the feature information belonging to the i-th type of preset label information.
The method for pedestrian re-identification according to claim 1, wherein the target image frame sequence and the label information are respectively re-identified in the pedestrian re-identification model completed by pre-training to complete the re-identification of the target image frame sequence and the label information. Before re-identification of pedestrians to be identified, including:

Collecting a first preset number of training samples, each of the training samples containing an image of a pedestrian to be identified and all preset label information corresponding to the pedestrian to be identified;

Use the training samples to train a pre-established machine learning model for training, to obtain a trained machine learning model;

Performing a model accuracy test on the machine learning model after training;

If the accuracy test of the machine learning model after the training is passed, it is determined that the machine learning model after the training is the pedestrian re-identification model;

If the accuracy test of the machine learning model after the training fails, increase the number of training samples, and return to perform training using the training sample training pre-established machine learning model to obtain the pedestrian Re-identify the model.
The method for pedestrian re-identification according to claim 4, wherein the training a pre-established machine learning model using the training samples to obtain the trained machine learning model comprises:

Use the pre-established machine learning model to re-identify all the preset label information corresponding to each pedestrian to be identified, obtain the probability that each pedestrian to be identified belongs to each type of preset label information, and determine each The preset label information with the highest probability corresponding to the pedestrian to be identified;

Respectively taking the preset label information with the highest probability corresponding to each of the pedestrians to be identified as the constraint conditions for training the machine learning model, and iterating the preset parameters of the machine learning model;

If the rate of change of the loss function value corresponding to the machine learning model tends to be stable, it is determined that the training of the machine learning model is completed, and the pedestrian re-identification model is obtained.
The method for pedestrian re-identification according to claim 4, wherein the loss function of the pedestrian re-identification model completed in the pre-training is:

among them,

Where N represents the total number of training samples, K represents the total number of categories of preset label information, p j represents the probability value of the current sample belonging to the j-th type of preset label information, y i is the true label information corresponding to the current sample, q i, j are the distribution ratios of p j , N sc represents the number of similar label information belonging to the current sample, and ε is a coefficient that balances the real label information and the similar label information,
Indicates that the feature information corresponding to the current sample belongs to the log probability value of the j-th type of preset label information,
Indicates that the feature information corresponding to the current sample belongs to the log probability value of the k-th type of preset label information,
Represents the first n tags whose output probability is greater than the preset probability threshold.
8. The method for pedestrian re-identification according to claim 6, wherein said performing a model accuracy test on the machine learning model after training comprises:

Input a second preset number of test samples into the machine learning model after training for analysis, and determine the rate of change of the loss function of the machine learning model after training;

If the change rate is less than or equal to the preset change rate threshold, it is determined that the test of the machine learning model after training passes;

If the rate of change is greater than the preset rate of change threshold, it is determined that the test of the machine learning model after the training fails.
A pedestrian re-identification device, which is characterized in that it comprises:

An acquiring module, configured to acquire a target image frame sequence from a pre-collected surveillance video stream, and the target image frame in the target image frame sequence contains image information of a pedestrian to be identified;

The first determining module is configured to identify the characteristic information of the pedestrian to be identified from the image information, and determine all tag information corresponding to the characteristic information;

The re-recognition module is used to perform pedestrian re-recognition on the target image frame sequence and all the tag information using the pre-trained pedestrian re-recognition model, and determine the target tag of the pedestrian to be recognized from all the tag information information;

The second determination module is configured to determine the re-identification result of the pedestrian to be identified based on the target tag information.
The pedestrian re-identification device according to claim 8, wherein the first determining module comprises:

A recognition unit, configured to use a feature information recognition model completed in advance to perform feature recognition on the pedestrian to be identified to obtain feature information of the pedestrian to be identified;

A calculation unit, configured to calculate the probability value of the feature information belonging to each type of preset label information;

The first determining unit is configured to: if the probability value of the feature information belonging to the first type of preset label information is greater than the probability value of belonging to the second type of preset label information, and the feature information belongs to the first type of preset If the probability value of the tag information is greater than the preset probability threshold, it is determined that the first type of preset tag information is the tag information corresponding to the feature information, and the second type of preset tag information is except for the first type of preset tag information. Set any type of preset label information except label information.
The pedestrian re-identification device according to claim 9, wherein the calculation unit comprises:

A preset probability normalization formula is used to calculate the probability value of the feature information belonging to each type of preset label information; the preset probability normalization formula is:

Wherein, p i represents the probability value of the feature information belonging to the i-th type of preset label information, and K represents the total number of types of preset label information,
Indicates the log probability value of the feature information belonging to the i-th type of preset label information.
A terminal includes a memory, a processor, and a computer program stored in the memory and running on the processor, wherein the processor implements the following steps when the processor executes the computer program:

Acquiring a target image frame sequence from a pre-collected surveillance video stream, where the target image frame in the target image frame sequence contains image information of a pedestrian to be identified;

Identifying characteristic information of the pedestrian to be identified from the image information, and determining all tag information corresponding to the characteristic information;

Using a pre-trained pedestrian re-recognition model to perform pedestrian re-recognition on the target image frame sequence and all the tag information, and determine the target tag information of the pedestrian to be identified from all the tag information;

The re-identification result of the pedestrian to be identified is determined based on the target tag information.
The terminal according to claim 11, wherein the identifying characteristic information of the pedestrian to be identified from the image information and determining the label information corresponding to the characteristic information comprises:

Using the pre-trained feature information recognition model to perform feature recognition on the pedestrian to be identified to obtain the feature information of the pedestrian to be identified;

Calculating the probability value of the feature information belonging to each type of preset label information;

If the probability value of the feature information belonging to the first type of preset tag information is greater than the probability value of belonging to the second type of preset tag information, and the probability value of the feature information belonging to the first type of preset tag information is greater than the preset The probability threshold of the first type is determined to be the tag information corresponding to the feature information, and the second type of preset tag information is any one other than the first type of preset tag information. Class preset label information.
The terminal according to claim 12, wherein the calculating the probability value of the characteristic information belonging to each type of preset label information comprises:

A preset probability normalization formula is used to calculate the probability value of the feature information belonging to each type of preset label information; the preset probability normalization formula is:

Wherein, p i represents the probability value of the feature information belonging to the i-th type of preset label information, and K represents the total number of types of preset label information,
Indicates the log probability value of the feature information belonging to the i-th type of preset label information.
The terminal according to claim 11, wherein the target image frame sequence and the label information are respectively re-identified in the pedestrian re-recognition model completed using the pre-training to complete the re-identification of the pedestrian to be identified. Before re-identification, including:

Collecting a first preset number of training samples, each of the training samples containing an image of a pedestrian to be identified and all preset label information corresponding to the pedestrian to be identified;

Use the training samples to train a pre-established machine learning model for training, to obtain a trained machine learning model;

Performing a model accuracy test on the machine learning model after training;

If the accuracy test of the machine learning model after the training is passed, it is determined that the machine learning model after the training is the pedestrian re-identification model;

If the accuracy test of the machine learning model after the training fails, increase the number of training samples, and return to perform training using the training sample training pre-established machine learning model to obtain the pedestrian Re-identify the model.
The terminal according to claim 14, wherein the training a pre-established machine learning model using the training samples to obtain a trained machine learning model comprises:

Use the pre-established machine learning model to re-identify all the preset label information corresponding to each pedestrian to be identified, obtain the probability that each pedestrian to be identified belongs to each type of preset label information, and determine each The preset label information with the highest probability corresponding to the pedestrian to be identified;

Respectively taking the preset label information with the highest probability corresponding to each of the pedestrians to be identified as the constraint conditions for training the machine learning model, and iterating the preset parameters of the machine learning model;

If the rate of change of the loss function value corresponding to the machine learning model tends to be stable, it is determined that the training of the machine learning model is completed, and the pedestrian re-identification model is obtained.
A computer-readable storage medium that stores a computer program, and is characterized in that, when the computer program is executed by a processor, the following steps are implemented:

Acquiring a target image frame sequence from a pre-collected surveillance video stream, where the target image frame in the target image frame sequence contains image information of a pedestrian to be identified;

Identifying characteristic information of the pedestrian to be identified from the image information, and determining all tag information corresponding to the characteristic information;

Using a pre-trained pedestrian re-recognition model to perform pedestrian re-recognition on the target image frame sequence and all the tag information, and determine the target tag information of the pedestrian to be identified from all the tag information;

The re-identification result of the pedestrian to be identified is determined based on the target tag information.
15. The computer-readable storage medium of claim 16, wherein the identifying characteristic information of the pedestrian to be identified from the image information and determining the label information corresponding to the characteristic information comprises:

Using the pre-trained feature information recognition model to perform feature recognition on the pedestrian to be identified to obtain the feature information of the pedestrian to be identified;

Calculating the probability value of the feature information belonging to each type of preset label information;

If the probability value of the feature information belonging to the first type of preset tag information is greater than the probability value of belonging to the second type of preset tag information, and the probability value of the feature information belonging to the first type of preset tag information is greater than the preset The probability threshold of the first type is determined to be the tag information corresponding to the feature information, and the second type of preset tag information is any one other than the first type of preset tag information. Class preset label information.
18. The computer-readable storage medium of claim 17, wherein the calculating the probability value of the characteristic information belonging to each type of preset label information comprises:

A preset probability normalization formula is used to calculate the probability value of the feature information belonging to each type of preset label information; the preset probability normalization formula is:

Wherein, p i represents the probability value of the feature information belonging to the i-th type of preset label information, and K represents the total number of types of preset label information,
It indicates that the feature information belongs to the log probability value of the i-th type of preset label information.
16. The computer-readable storage medium of claim 16, wherein the target image frame sequence and the tag information are re-identified in the pedestrian re-identification model completed in advance to complete the re-identification of the target image frame sequence and the label information. Before re-identification of pedestrians to be identified, including:

Collecting a first preset number of training samples, each of the training samples containing an image of a pedestrian to be identified and all preset label information corresponding to the pedestrian to be identified;

Use the training samples to train a pre-established machine learning model for training, to obtain a trained machine learning model;

Performing a model accuracy test on the machine learning model after training;

If the accuracy test of the machine learning model after the training is passed, it is determined that the machine learning model after the training is the pedestrian re-identification model;

If the accuracy test of the machine learning model after the training fails, increase the number of training samples, and return to perform training using the training sample training pre-established machine learning model to obtain the pedestrian Re-identify the model.
19. The computer-readable storage medium of claim 19, wherein the training a pre-established machine learning model using the training samples to obtain the trained machine learning model comprises:

Use the pre-established machine learning model to re-identify all the preset label information corresponding to each pedestrian to be identified, obtain the probability that each pedestrian to be identified belongs to each type of preset label information, and determine each The preset label information with the highest probability corresponding to the pedestrian to be identified;

Respectively taking the preset label information with the highest probability corresponding to each of the pedestrians to be identified as the constraint conditions for training the machine learning model, and iterating the preset parameters of the machine learning model;

If the rate of change of the loss function value corresponding to the machine learning model tends to be stable, it is determined that the training of the machine learning model is completed, and the pedestrian re-identification model is obtained.