Urban rail transit passenger re-identification method and system
Through the integration of deep learning and machine learning models, combined with automatic image annotation tools, a dynamic passenger feature library is established, which solves the problem of decreasing recognition accuracy caused by passenger appearance changes in the existing technology, and achieves higher recognition accuracy and adaptability.
Patent Information
- Application Number
- CN202411519549.1
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-10-29
- Publication Date
- 2025-02-28
AI Technical Summary
The existing urban rail transit passenger re-identification technology is difficult to maintain high accuracy when passenger appearance changes, and cannot meet the requirements of invisible payment.
A fusion method of deep learning model and machine learning model is adopted, combined with automatic image annotation tool, a dynamic passenger feature image library is established, and the recognition model is updated through incremental learning, and the weight of the feature vector is dynamically adjusted to adapt to the changes in passenger flow density during different time periods.
It improves the accuracy and robustness of passenger re-identification, can more effectively adapt to passenger appearance changes, and meets the high accuracy requirements of invisible payment.
Smart Images
Figure CN119540988A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of target recognition, and in particular to a method and system for re-recognizing passengers in urban rail transit. Background Art
[0002] Daily commuting is an important part of riding urban rail transit. In order to accurately identify each passenger, the image recognition algorithm cannot rely solely on face recognition, but must take into account multiple factors, such as gender, age, height, weight, etc. The appearance of each passenger cannot be guaranteed to be the same when riding the subway every day. Therefore, if a fixed model is used to complete the re-identification of passengers, the recognition accuracy cannot meet the requirements of contactless payment. Existing passenger re-identification basically uses the same set of feature images of a passenger as a reference for identification, which results in low recognition accuracy every time the passenger's appearance, clothing, etc. change. Summary of the invention
[0003] The object of the present invention is to provide a method and system for re-identifying passengers in urban rail transit, so as to solve at least one technical problem existing in the above-mentioned background technology.
[0004] In order to achieve the above object, the present invention adopts the following technical solutions:
[0005] In a first aspect, the present invention provides a method for training a passenger re-identification model for urban rail transit, comprising:
[0006] Establish a rail transit passenger feature image library, where each passenger corresponds to a feature image set;
[0007] Use the recognition model to re-identify the passenger in each frame, update the passenger's feature image set, and append the feature map obtained after recognition to the passenger's feature image set;
[0008] Perform snapshot operations at the same time interval to obtain the current passenger and the corresponding feature image set;
[0009] For each passenger's feature picture set in the feature picture set, an automatic image annotation tool is used to annotate the image, annotate the passenger's identity features, and merge the annotated feature picture set with the original passenger feature picture set to obtain an incremental training data set; the incremental training data set is used to perform incremental learning on the recognition model to obtain an updated model as the urban rail transit passenger re-identification model.
[0010] Furthermore, the recognition model is used to re-identify passengers for each frame of image as follows: passenger re-identification based on the fusion of deep learning model and machine learning model includes the following processes: weight allocation of feature vectors based on changes in passenger flow density in different time periods; the obtained feature vectors with weight allocation based on passenger flow density in different time periods are input into the traditional machine learning model, the deep learning model classifies the feature vectors, and weighted averages the classification prediction results according to the actual situation.
[0011] Furthermore, different weights are assigned to the feature vectors based on the passenger flow density change factors in different time periods, including: assuming that the feature vector is It is a 2×1 matrix; where x1,x2,…x n Represents local feature elements, y1,y2,…y n Represents the global feature element, corresponding to each feature x i ,y i The initial weight is Calculate the passenger flow density in different time periods; for each layout feature x i , define the weight as w i (d), the weight calculation formula is: For each global feature y j , define the weight as w j (d), the weight calculation formula is: in Respectively represent the initial weights of each local feature and each global feature element, β represents the weight adjustment coefficient in practical application. The determination of the coefficient is related to the optimal value obtained by repeated training of the passenger recognition model. D0 represents the average passenger flow density in the calculation area. The weighted feature vector is calculated as:
[0012]
[0013] Among them, · indicates that each element in the matrix is multiplied.
[0014] Furthermore, the deep learning model is used to classify the passengers on the feature vectors, and the classification prediction results are weighted averaged according to the actual situation, including: the feature vector obtained by the deep learning model is X′1, and the feature vector processed by the traditional machine learning model is X′2; the above feature vectors are concatenated to obtain X combine =X′1⊕X′2; combine Input into the traditional machine learning model random forest, each input will get a prediction result, let the total prediction result be X r1 ′、X r2 ′…X rn′, perform weighted fusion on the prediction results to obtain in The final passenger re-identification result is obtained based on the fused prediction results.
[0015] Furthermore, the initial weight value determination includes: collecting multiple passenger images at different time periods and locations; using the random forest ensemble learning algorithm, by constructing multiple decision trees, counting the number of times b that each feature in the image is split in all decision trees, and the total number of splits of all features is B, then the initial weight of each feature is
[0016] In a second aspect, the present invention provides an urban rail transit passenger re-identification model training system, comprising:
[0017] Establish a module for establishing a rail transit passenger feature image library, where each passenger corresponds to a feature image set; use the recognition model to re-identify the passenger for each frame of the image, update the passenger's feature image set, and append the feature image obtained after identification to the passenger's feature image set; perform snapshot operations at the same time interval to obtain the current passenger and the corresponding feature image set;
[0018] The training module is used to use an automatic image annotation tool to annotate the feature picture set of each passenger in the feature picture set, annotate the identity characteristics of the passengers, merge the annotated feature picture set with the original passenger feature picture set to obtain an incremental training data set; use the incremental training data set to perform incremental learning on the recognition model to obtain an updated model as the urban rail transit passenger re-identification model.
[0019] In a third aspect, the present invention provides a method for re-identifying a passenger in urban rail transit, comprising:
[0020] Get the picture to be identified;
[0021] The acquired picture to be identified is processed using the urban rail transit passenger re-identification model to obtain a pedestrian re-identification result in the picture to be identified; wherein the urban rail transit passenger re-identification model is trained using the urban rail transit passenger re-identification model training method as described in the first aspect.
[0022] In a fourth aspect, the present invention provides an urban rail transit passenger re-identification system, comprising:
[0023] An acquisition module is used to acquire the image to be identified;
[0024] A processing module is used to process the acquired picture to be identified using the urban rail transit passenger re-identification model to obtain a pedestrian re-identification result in the picture to be identified; wherein the urban rail transit passenger re-identification model is trained using the urban rail transit passenger re-identification model training method as described in the first aspect.
[0025] In a fifth aspect, the present invention provides a non-transitory computer-readable storage medium, which is used to store computer instructions. When the computer instructions are executed by a processor, the urban rail transit passenger re-identification method as described in the third aspect is implemented.
[0026] In a sixth aspect, the present invention provides a computer device comprising a memory and a processor, wherein the processor and the memory communicate with each other, the memory stores program instructions executable by the processor, and the processor calls the program instructions to execute the urban rail transit passenger re-identification method as described in the third aspect.
[0027] In the seventh aspect, the present invention provides an electronic device, comprising: a processor, a memory and a computer program; wherein the processor is connected to the memory, the computer program is stored in the memory, and when the electronic device is running, the processor executes the computer program stored in the memory so that the electronic device executes instructions for implementing the urban rail transit passenger re-identification method as described in the third aspect.
[0028] The beneficial effects of the present invention are as follows: a rail transit passenger feature image library is established in view of the diverse appearance features of passengers, and a dynamic model training method is implemented according to the operating characteristics of rail transit, so as to continuously optimize the accuracy of passenger re-identification and improve the accuracy rate of passenger re-identification.
[0029] Additional advantages of the present invention will be more clearly given in the following description or learned through the practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS
[0030] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative work.
[0031] Figure 1 This is a flow chart of establishing a rail transit passenger feature image library according to an embodiment of the present invention. DETAILED DESCRIPTION
[0032] The embodiments of the present invention are described in detail below, and examples of the embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements with the same or similar functions. The embodiments described below by the accompanying drawings are exemplary and are only used to explain the present invention, and cannot be interpreted as limiting the present invention.
[0033] It should be understood by those skilled in the art that unless otherwise defined, all terms (including technical and scientific terms) used herein have the same meaning as commonly understood by those skilled in the art to which this invention belongs.
[0034] It should also be understood that terms, such as those defined in commonly used dictionaries, should be understood to have a meaning consistent with that in the context of the prior art and will not be interpreted in an idealized or overly formal sense unless as defined herein.
[0035] Those skilled in the art will appreciate that, unless otherwise stated, the singular forms "a", "an", "said" and "the" used herein may also include plural forms. It should be further understood that the term "comprising" used in the specification of the present invention refers to the presence of the features, integers, steps, operations, elements and / or components, but does not exclude the presence or addition of one or more other features, integers, steps, operations, elements and / or groups thereof.
[0036] In the description of this specification, the description with reference to the terms "one embodiment", "some embodiments", "example", "specific example", or "some examples" means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present invention. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or more embodiments or examples in a suitable manner. Different embodiments or examples described in this specification and features of different embodiments or examples may be combined and combined by those skilled in the art without contradiction.
[0037] To facilitate understanding of the present invention, the present invention is further explained below with reference to specific embodiments in conjunction with the accompanying drawings, and the specific embodiments do not constitute a limitation on the embodiments of the present invention.
[0038] Those skilled in the art should understand that the drawings are merely schematic diagrams of embodiments, and the components in the drawings are not necessarily necessary for implementing the present invention.
[0039] The present invention provides a method for re-identifying passengers in urban rail transit. The method establishes a rail transit passenger feature image library based on the diverse appearance features of passengers, implements a dynamic model training method according to the operation characteristics of rail transit, continuously optimizes the accuracy of passenger re-identification, and improves the accuracy rate of passenger re-identification.
[0040] Example 1
[0041] In this embodiment 1, first, a system for training a model for re-identifying a passenger in urban rail transit is provided, and the system includes: an establishment module, which is used to establish a rail transit passenger feature image library, and each passenger corresponds to a feature image set; using the recognition model to re-identify the passenger for each frame of the image, updating the feature image set of the passenger, and appending the feature image obtained after the recognition to the feature image set of the passenger; taking snapshots at the same time interval to obtain the current passenger and the corresponding feature image set; a training module, which is used to annotate the feature image set of each passenger in the feature image set using an automatic image annotation tool, annotate the identity characteristics of the passenger, merge the annotated feature image set with the original passenger feature image set, and obtain an incremental training data set; using the incremental training data set to perform incremental learning on the recognition model, and obtain an updated model as the urban rail transit passenger re-identification model.
[0042] In this embodiment 1, the above-mentioned system is used to implement a method for training a passenger re-identification model for urban rail transit, which includes: establishing a rail transit passenger feature image library, where each passenger corresponds to a feature image set; using a recognition model to re-identify the passenger for each frame of the image, updating the passenger's feature image set, and appending the feature image obtained after recognition to the passenger's feature image set; performing snapshot operations at the same time interval to obtain the current passenger and the corresponding feature image set; using an automatic image annotation tool to annotate the feature image set of each passenger in the feature image set, annotating the passenger's identity features, and merging the annotated feature image set with the original passenger feature image set to obtain an incremental training data set; using the incremental training data set to incrementally learn the recognition model to obtain an updated model as the urban rail transit passenger re-identification model.
[0043] In this embodiment, the identity characteristics of the passenger include gender, height, age and body shape.
[0044] In this embodiment, a snapshot operation is performed at a specified time T0 every day to obtain the current passengers and their corresponding feature picture sets.
[0045] In this embodiment, a fusion method of machine learning and deep learning model is used to perform passenger re-identification. A deep learning model is used to extract image features, wherein passenger flow density in different time periods has different effects on feature extraction, and the extracted features are used to train a machine learning model for classification and recognition.
[0046] Specifically, in this embodiment, the weight allocation of feature vectors is performed based on the changes in passenger flow density in different time periods, including: In the passenger re-identification model, the feature vector is a mathematical representation used to describe the key features of the passenger. Each element in the feature vector represents a feature of the passenger. The present invention comprehensively considers multiple factors, such as gender, age, height, body shape, etc., and comprehensively analyzes and synthesizes the passenger features through a convolutional neural network. After multi-layer convolution and pooling operations, feature vectors representing the local features (face, hairstyle and hair color) and global features (clothing, color, body shape) of the passenger are output. After the feature vector is extracted, the passenger is re-identified by training the machine learning model. In this embodiment, weights are allocated to different elements in the feature vector according to the changes in passenger flow density in different time periods. Feature elements under different weights mean that when performing passenger re-identification, more emphasis is placed on the comparison of different features. For example, in the feature vector containing the above features, if the passenger flow is large during the morning peak period, the mutual occlusion between passengers makes it difficult to accurately capture and identify the local detail features, while the overall contour and color features will not be completely blocked, and the capture and identification are relatively stable. Based on the above situation, all the features in the feature extraction vector are given higher weights, and the classifier model will focus more on the comparison of global features when classifying passengers.
[0047] The specific steps for assigning different weights to feature vectors based on the passenger flow density change factors in different time periods are as follows:
[0048] Step 1: Let the eigenvector be It is a 2×1 matrix. n Represents local feature elements, y1,y2,…y n Represents the global feature element. Corresponding to each feature x i ,y i The initial weight is The initial weight value determination includes: collecting multiple passenger images at different time periods and locations, using the random forest ensemble learning algorithm, constructing multiple decision trees, and counting the number of times each feature is used for splitting in all decision trees. For example, if the facial feature is used b times in the decision splitting tree, and the total number of splits of all features is B, then the initial weight of the facial feature is
[0049] Step 2: Calculate the passenger flow density in different time periods. Divide the time into morning peak (8:00-10:00), flat peak (11:00-16:00, 22:00-0:00) and evening peak (17:00-21:00). Based on the video data in the subway station, count the number of passengers in different areas at the current time. Suppose the number of passengers in a certain area is Q and the area is A. The passenger flow density at the current time
[0050] Step 3: For each layout feature x i , define the weight as w i (d), the weight calculation formula is: For each global feature y j , define the weight as w j (d), the weight calculation formula is: in Respectively represent the initial weights of each local feature and each global feature element, β represents the weight adjustment coefficient in practical applications, and the determination of the coefficient is related to the optimal value obtained when the passenger recognition model is repeatedly trained. D0 represents the average passenger flow density in the calculation area.
[0051] Step 4: Calculate the weighted eigenvector:
[0052]
[0053] The calculation in the formula is defined as multiplying each element in the matrix, that is, x′ i =w i (d)·x i , y′ j =w j (d) y j
[0054]
[0055] In this embodiment, based on the changes in passenger flow density in different time periods, when using a convolutional neural network to calculate the feature vector, the impact of the passenger flow density factor on the passenger re-identification model is added, and different weights are assigned to the feature vectors according to different regional densities. For example, during peak hours in the morning and evening, the weight of the feature vector can be adjusted according to the high passenger flow density, and more emphasis is placed on easily observable global passenger features to improve the model's recognition and robustness.
[0056] Passenger re-identification based on the fusion of deep learning model and machine learning model includes the following processes:
[0057] Deep learning models have advantages in processing large-scale data and a large number of passenger feature images. Traditional machine learning models can perform faster passenger re-identification based on the input feature vectors, thereby improving the recognition efficiency of target passengers. The feature vectors obtained above with weights assigned based on passenger flow density in different time periods are input into the traditional machine learning model. The deep learning model classifies the feature vectors and performs weighted average on the classification prediction results based on the actual situation. The specific implementation steps are as follows:
[0058] Step 1: The feature vector obtained by the deep learning model is X′1, and the feature vector processed by the traditional machine learning model is X′2;
[0059] Step 2: Concatenate the above feature vectors to get X combine =X′1⊕X′2;
[0060] Step 3: Place the X combine Input into the traditional machine learning model random forest, each input will get a prediction result, for example, the corresponding prediction result of facial features is X r1 ′, let the total prediction result be X r1 ′、X r2 ′…X rn ′, weighted fusion of the prediction results, and get in
[0061] Step 4: Obtain the final passenger re-identification result based on the fused prediction results.
[0062] Example 2
[0063] In this embodiment 2, firstly, a system for re-identifying a passenger in urban rail transit is provided, which includes: an acquisition module for acquiring a picture to be identified; and a processing module for processing the acquired picture to be identified using a model for re-identifying a passenger in urban rail transit to obtain a result of re-identifying a pedestrian in the picture to be identified.
[0064] In this embodiment 2, a method for urban rail transit passenger re-identification is implemented using the above-mentioned system. The implementation process of the method includes: using an acquisition module to acquire a picture to be identified; using a processing module to process the acquired picture to be identified using an urban rail transit passenger re-identification model to obtain a pedestrian re-identification result in the picture to be identified.
[0065] In this embodiment 2, the urban rail transit passenger re-identification model is trained using the urban rail transit passenger re-identification model training method described in embodiment 1.
[0066] Example 3
[0067] In this embodiment 3, firstly, a system for re-identifying a passenger in urban rail transit is provided, which includes: an acquisition module for acquiring a picture to be identified; a processing module for processing the acquired picture to be identified using an urban rail transit passenger re-identification model to obtain a result of re-identifying a pedestrian in the picture to be identified.
[0068] In this embodiment 3, a method for urban rail transit passenger re-identification is implemented using the above-mentioned system. The implementation process of the method includes: using an acquisition module to acquire a picture to be identified; using a processing module to process the acquired picture to be identified using an urban rail transit passenger re-identification model to obtain a pedestrian re-identification result in the picture to be identified.
[0069] In this embodiment 3, the training process of the urban rail transit passenger re-identification model includes:
[0070] Step 1: Establish a rail transit passenger feature image library. Each passenger corresponds to a feature image set, such as Figure 1 As shown; Step 2: Use the existing model to re-identify passengers; Step 3: Loop through the passenger re-identification operation, re-identify passengers for each frame of image, update the passenger feature image list, and add the feature images obtained after identification to the passenger feature image list; Step 4: Perform a snapshot operation, perform a snapshot operation at a specified time T0 every day, and obtain the current passenger and its corresponding feature image set S; Step 5: Image annotation and model training (incremental learning), first, for each passenger feature image set in the snapshot S, use the automatic image annotation tool to annotate the image, annotate the passenger's gender, age, height, body shape and other features; secondly, merge the annotated feature image set with the original passenger feature image set to obtain an incremental training data set; finally, use the incremental training data set to perform incremental learning on the existing model M1. Among them, the incremental support vector machine is used to perform incremental learning on the existing model M1. After the incremental learning is completed, the updated model M2 is obtained.
[0071] Update the model and continue the re-identification work. Use the new model M2 to replace the original model M1, which becomes the model for the next passenger re-identification, and continue to perform the passenger re-identification work.
[0072] Example 4
[0073] This embodiment 4 provides a non-transitory computer-readable storage medium, which is used to store computer instructions. When the computer instructions are executed by the processor, the urban rail transit passenger re-identification method as described above is implemented, and the method includes: using an acquisition module to acquire a picture to be identified; using a processing module to process the acquired picture to be identified using an urban rail transit passenger re-identification model to obtain a pedestrian re-identification result in the picture to be identified. Among them, training the rail transit passenger re-identification model includes: establishing a rail transit passenger feature image library, each passenger corresponds to a feature image set; using a recognition model to re-identify the passenger for each frame of the image, updating the passenger's feature image set, and appending the feature map obtained after identification to the passenger's feature image set; performing snapshot operations at the same time interval to obtain the current passenger and the corresponding feature image set; for each passenger's feature image set in the feature image set, using an automatic image annotation tool to annotate the image, annotate the passenger's identity features, merge the annotated feature image set with the original passenger feature image set, and obtain an incremental training data set; using the incremental training data set to incrementally learn the recognition model, and obtain an updated model as the urban rail transit passenger re-identification model.
[0074] Example 5
[0075] This embodiment 5 provides a computer device, including a memory and a processor, the processor and the memory communicate with each other, the memory stores program instructions that can be executed by the processor, and the processor calls the program instructions to execute the above-mentioned urban rail transit passenger re-identification method, the method comprising: using an acquisition module to acquire a picture to be identified; using a processing module to process the acquired picture to be identified using an urban rail transit passenger re-identification model to obtain a pedestrian re-identification result in the picture to be identified. Among them, training the rail transit passenger re-identification model includes: establishing a rail transit passenger feature image library, each passenger corresponds to a feature image set; using a recognition model to re-identify the passenger for each frame of the image, updating the passenger's feature image set, and appending the feature map obtained after identification to the passenger's feature image set; performing a snapshot operation at the same time interval to obtain the current passenger and the corresponding feature image set; for each passenger's feature image set in the feature image set, using an automatic image annotation tool to annotate the image, annotate the passenger's identity features, merge the annotated feature image set with the original passenger feature image set, and obtain an incremental training data set; using the incremental training data set to incrementally learn the recognition model, and obtain an updated model as the urban rail transit passenger re-identification model.
[0076] Example 6
[0077] This embodiment 6 provides an electronic device, including: a processor, a memory and a computer program; wherein the processor is connected to the memory, and the computer program is stored in the memory. When the electronic device is running, the processor executes the computer program stored in the memory so that the electronic device executes instructions for implementing the urban rail transit passenger re-identification method as described above, the method comprising: using an acquisition module to acquire a picture to be identified; using a processing module to process the acquired picture to be identified using an urban rail transit passenger re-identification model to obtain a pedestrian re-identification result in the picture to be identified. Among them, training the rail transit passenger re-identification model includes: establishing a rail transit passenger feature image library, each passenger corresponds to a feature image set; using the recognition model to re-identify the passenger for each frame of the image, updating the passenger's feature image set, and appending the feature image obtained after recognition to the passenger's feature image set; performing snapshot operations at the same time interval to obtain the current passenger and the corresponding feature image set; for the feature image set of each passenger in the feature image set, using an automatic image annotation tool to annotate the image, annotate the passenger's identity features, merge the annotated feature image set with the original passenger feature image set, and obtain an incremental training data set; using the incremental training data set to perform incremental learning on the recognition model, and obtain an updated model as the urban rail transit passenger re-identification model.
[0078] Those skilled in the art will appreciate that embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.
[0079] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.
[0080] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.
[0081] These computer program instructions can also be loaded onto a computer or other programmable data processing device, and a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, so that the instructions executed on the computer or other programmable device provide for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.
[0082] Although the above describes the specific implementation mode of the present invention in conjunction with the accompanying drawings, it is not intended to limit the scope of protection of the present invention. Those skilled in the art should understand that various modifications or variations that can be made by those skilled in the art without creative work on the basis of the technical solution disclosed in the present invention should be included in the scope of protection of the present invention.
Claims
1. A method for training a passenger re-identification model for urban rail transit, characterized in that: include: Establish a rail transit passenger feature image library, where each passenger corresponds to a feature image set; Use the recognition model to re-identify the passenger in each frame, update the passenger's feature image set, and append the feature map obtained after recognition to the passenger's feature image set; Perform snapshot operations at the same time interval to obtain the current passenger and the corresponding feature image set; For each passenger's feature picture set in the feature picture set, an automatic image annotation tool is used to annotate the image, annotate the passenger's identity features, and merge the annotated feature picture set with the original passenger feature picture set to obtain an incremental training data set; the incremental training data set is used to perform incremental learning on the recognition model to obtain an updated model as the urban rail transit passenger re-identification model.
2. The urban rail transit passenger re-identification model training method according to claim 1 is characterized in that: The recognition model is used to re-identify passengers for each frame of image: Passenger re-identification based on the fusion of deep learning model and machine learning model includes the following processes: weight allocation of feature vectors based on changes in passenger flow density in different time periods; the obtained feature vectors with weight allocation based on passenger flow density in different time periods are input into the traditional machine learning model, and the deep learning model classifies the feature vectors and performs weighted average of the classification prediction results according to the actual situation.
3. The urban rail transit passenger re-identification model training method according to claim 2 is characterized in that: Different weights are assigned to the feature vectors based on the passenger flow density change factors in different time periods, including: assuming that the feature vector is It is a 2×1 matrix; where x1,x2,…x n Represents local feature elements, y1,y2,…y n Represents the global feature element, corresponding to each feature x i ,y i The initial weight is Calculate the passenger flow density in different time periods; for each layout feature x i , define the weight as w i (d), the weight calculation formula is: For each global feature y j , define the weight as w j (d), the weight calculation formula is: in Respectively represent the initial weights of each local feature and each global feature element, β represents the weight adjustment coefficient in practical application. The determination of the coefficient is related to the optimal value obtained by repeated training of the passenger recognition model. D0 represents the average passenger flow density in the calculation area. The weighted feature vector is calculated as: Among them, · indicates that each element in the matrix is multiplied.
4. The urban rail transit passenger re-identification model training method according to claim 3 is characterized in that: The deep learning model classifies the feature vectors for passengers, and performs weighted average on the classification prediction results according to the actual situation, including: the feature vector obtained by the deep learning model is X′1, and the feature vector processed by the traditional machine learning model is X′2; the above feature vectors are concatenated to obtain X combine Input into the traditional machine learning model random forest, each input will get a prediction result, let the total prediction result be X r1 ′、X r2 ′…X rn ′, perform weighted fusion on the prediction results to obtain in The final passenger re-identification result is obtained based on the fused prediction results.
5. The urban rail transit passenger re-identification model training method according to claim 3 is characterized in that: The initial weight value determination includes: collecting multiple passenger images at different time periods and locations; using the random forest ensemble learning algorithm, by constructing multiple decision trees, counting the number of times b that each feature in the image is split in all decision trees, and the total number of splits of all features is B, then the initial weight of each feature is 6. An urban rail transit passenger re-identification model training system, characterized in that: include: Establishing a module for establishing a rail transit passenger feature image library, where each passenger corresponds to a feature image set; Use the recognition model to re-identify the passenger in each frame, update the passenger's feature image set, and append the feature map obtained after recognition to the passenger's feature image set; perform snapshot operations at the same time interval to obtain the current passenger and the corresponding feature image set; The training module is used to use an automatic image annotation tool to annotate the feature picture set of each passenger in the feature picture set, annotate the identity characteristics of the passengers, merge the annotated feature picture set with the original passenger feature picture set to obtain an incremental training data set; use the incremental training data set to perform incremental learning on the recognition model to obtain an updated model as the urban rail transit passenger re-identification model.
7. A method for re-identifying passengers in urban rail transit, characterized in that: include: Get the picture to be identified; The acquired picture to be identified is processed using an urban rail transit passenger re-identification model to obtain a pedestrian re-identification result in the picture to be identified; wherein the urban rail transit passenger re-identification model is trained using the urban rail transit passenger re-identification model training method according to any one of claims 1 to 5.
8. An urban rail transit passenger re-identification system, characterized in that: include: An acquisition module is used to acquire the image to be identified; A processing module is used to process the acquired picture to be identified by using the urban rail transit passenger re-identification model to obtain a pedestrian re-identification result in the picture to be identified; wherein the urban rail transit passenger re-identification model is trained by using the urban rail transit passenger re-identification model training method according to any one of claims 1 to 5.
9. A computer device, characterized in that: It includes a memory and a processor, the processor and the memory communicate with each other, the memory stores program instructions that can be executed by the processor, and the processor calls the program instructions to execute the urban rail transit passenger re-identification method as described in claim 7.
10. An electronic device, characterized in that: include: A processor, a memory and a computer program; wherein the processor is connected to the memory, the computer program is stored in the memory, and when the electronic device is running, the processor executes the computer program stored in the memory so that the electronic device executes instructions for implementing the urban rail transit passenger re-identification method as described in claim 7.