Multi-granularity feature pedestrian re-identification method based on attitude guidance

Abstract

The invention belongs to the field of computer vision, discloses a multi-granularity feature pedestrian re-identification method based on attitude guidance, and aims to solve the problem of low identification precision of existing pedestrian re-identification. The method comprises the following steps: (1) constructing a multi-granularity feature extraction network model; (2) training the multi-granularity feature extraction network model to obtain a trained multi-granularity feature extraction network model; and (3) inputting a pedestrian image to be identified into the trained multi-granularity feature extraction network model, fusing global features, local medium-granularity features and local fine-granularity features, and further performing multi-granularity pedestrian re-identification. The invention constructs a multi-granularity pedestrian re-identification method based on global feature and local feature fusion of attitude guidance, and effectively solves the problems that pedestrian key information is difficult to extract in a disordered scene and the identification precision is not high due to the failure of a global feature method during local shielding.

Description

Technical Field

[0001] The present invention belongs to the field of computer vision, and in particular relates to a multi-granularity feature pedestrian re-identification method based on posture guidance. Background Art

[0002] Pedestrian re-identification aims to associate images of a pedestrian captured at different locations and by different cameras in order to retrieve specific pedestrians across surveillance images and device videos. As this technology is widely used in fields such as intelligent video surveillance and large-scale personnel tracking, more and more challenging problems have been exposed. The main challenges include but are not limited to background clutter, perspective changes, different lighting conditions, and partial occlusion. Especially in actual application scenarios, such as crowded common places or in the presence of obstacles, some parts of the pedestrian's body may be completely or partially occluded, which makes it difficult for traditional global feature-based methods to accurately extract complete information. Therefore, how to improve feature recognition capabilities in pedestrian images captured in real life scenarios is the primary task of current ReID research.

[0003] The advent of the era of deep learning has produced many networks with powerful learning and feature extraction capabilities. In the early stages of its development, the purpose of most methods was to obtain significant information from the overall image of pedestrians in order to obtain the global feature expression of pedestrian images. However, if part of the pedestrian's body is blocked or there is a detection error, the camera will not be able to capture the entire pedestrian, and the global feature expression is likely to ignore some key local information, affecting the performance of the global feature and resulting in a decrease in recognition accuracy. Therefore, the introduction of more complex local features has become a research hotspot.

[0004] With the further development of technology, image segmentation, human posture key point alignment and instance segmentation have become common methods for extracting local features. However, the method of segmenting pedestrian images can easily lead to the lack of context between different parts of the human body. These methods often ignore the natural connection between different parts of the human body, resulting in the lack of contextual information, which in turn affects the effectiveness and robustness of the features. In addition, the existing real segmentation methods roughly segment the background and often cannot generalize the ReID real scene well.

[0005] At the same time, in order to better improve the anti-interference ability of the network and extract more discriminative features, in recent years, the method of introducing external clues has been used in pedestrian re-identification systems. Through the posture estimation method, key point areas are obtained and the classification information areas are focused on. However, although the current method of introducing posture estimation can extract local features of pedestrians under complex backgrounds, it fails to fully explore the potential of posture information in enhancing local feature expression and understanding the overall structure of pedestrians.

[0006] Therefore, in order to achieve robust pedestrian re-identification, it is crucial to make full use of the information of human key points and allow deep neural networks to build more powerful feature representations in limited input data. Summary of the invention

[0007] The present invention provides a multi-granularity feature pedestrian re-identification method based on posture guidance, which makes full use of the pedestrian key point coordinates and key point confidence information obtained by the pedestrian key point detection algorithm, and constructs a multi-granularity pedestrian re-identification method based on the fusion of global features and local features guided by posture, which effectively solves the problems of difficulty in extracting pedestrian key information in cluttered scenes and low recognition accuracy caused by failure of the global feature method when there is local occlusion.

[0008] In order to solve the technical problem, the technical solution adopted by the present invention is:

[0009] A multi-granularity feature pedestrian re-identification method based on posture guidance, characterized by comprising:

[0010] (1) constructing a multi-granularity feature extraction network model, wherein the multi-granularity feature extraction network model includes a backbone feature extraction network, a human key point extractor, a global branch network, a local medium-granularity feature branch, and a local fine-grained graph convolution branch;

[0011] (2) training the multi-granularity feature extraction network model to obtain a trained multi-granularity feature extraction network model;

[0012] (3) The image of the pedestrian to be identified is input into the trained multi-granularity feature extraction network model, and the global features, local medium-granularity features and local fine-granularity features are integrated to perform multi-granularity pedestrian re-identification.

[0013] In some embodiments, the pedestrian image to be identified is preprocessed before being input into the trained multi-granularity feature extraction network model in step (3), and the preprocessing includes using four data enhancement methods: random erasing, random padding, random cropping, and horizontal flipping, so as to improve the robustness of the model to occluded pedestrian images and prevent overfitting.

[0014] In some embodiments, the backbone feature extraction network is composed of a multi-layer convolutional neural network, which is responsible for extracting backbone network features from input pedestrian images during training and testing; the human key point extractor is composed of a multi-layer convolutional neural network, which is responsible for extracting key point information from input pedestrian images during training and testing, wherein the key point information includes key point coordinates, key point confidence, and key point heat map.

[0015] In some embodiments, the global branch network combines the backbone network features extracted by the backbone feature extraction network with the corresponding key point heat map as input and performs Hadamard product to obtain global features, and then outputs the global features as coarse-grained features through a global average pooling layer and a fully connected classification layer.

[0016] In some embodiments, the local medium-granularity feature branch uses the key point coordinates extracted by the human key point extractor and divides the body into several regions according to the actual posture of the human body, and then extracts several body regions in the pedestrian image according to the key point coordinates, and then passes the several body regions through the global average pooling layer and the fully connected classification layer as medium-granularity feature outputs.

[0017] In some embodiments, the local medium-granularity feature branch is divided into three regions according to the key point coordinates and the actual posture of the human body, and the head region, upper body region and lower body region are extracted, and the head region overlaps with the upper body region, and the upper body region overlaps with the lower body region.

[0018] In some embodiments, the local medium-granularity feature branch generates a corresponding bounding box B according to the key point position of each body region (head region, upper body region and lower body region). j ; where j is 1, 2, 3, representing top, middle, and bottom respectively; bounding box B j The coordinates of the upper left corner and lower right corner are calculated according to the key point position as follows:

[0019] B j =[min(x j )-5,min(y j )-5,max(x j ),max(y j )];

[0020] Among them, x j and j Represents the horizontal and vertical coordinates of the key point of the jth region.

[0021] In some embodiments, the local medium-granularity feature branch also uses the key point confidence S i , and score the visibility of each body region; for the i-th body region, its visibility score is calculated according to the following formula:

[0022]

[0023] Among them, S i is the confidence of the key points in the body area, and N is the number of key points.

[0024] In some embodiments, the local medium-granularity feature branch processes each pedestrian image to be identified and the candidate image, identifies the unobstructed shared feature area in the two images, and calculates the distance between the features based on the unobstructed shared feature area. The calculation formula is as follows:

[0025]

[0026] Where D is the cosine distance function, represents the alignment features of the i-th part of the query set image, and accordingly, represents the i-th part of the candidate set image, and dist represents the final total distance.

[0027] In some embodiments, the local fine-grained graph convolution branch combines the backbone network features extracted by the backbone feature extraction network with the corresponding key point heat map to obtain a number of key point features, and inputs the combined key point features as graph nodes, and then enters the multi-layer graph convolution network to obtain a number of learned key point features; then the learned key point features are output as fine-grained features through the global average pooling layer and the fully connected classification layer.

[0028] Compared with the prior art, the present invention has the following beneficial effects:

[0029] The multi-granularity feature pedestrian re-identification method based on posture guidance of the present invention utilizes the posture guidance network to progressively learn pedestrian features from coarse granularity to fine granularity, thereby focusing on key information in a multi-granularity manner, enabling the network to better focus on the significant features of pedestrians in a cluttered background and reduce the interference of irrelevant information.

[0030] The present invention extracts diverse features of different granularities through local feature branches and local graph convolution branches, effectively making up for the defect of global information failure during occlusion, and improves the generalization ability of the network by forcing the network to pay attention to the correlation between features and the remaining area.

[0031] The present invention fuses coarse-grained global features with multi-level fine-grained local features to obtain more discriminative multi-granular features, so that candidate pedestrian images can be more accurately identified even if important information is lost in the pedestrian image.

[0032] The recognition effects of the present invention are higher than the evaluation indicators of the current mainstream models and have a very high recognition accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] Figure 1 It is a structural diagram of a multi-granularity feature extraction network model according to an embodiment of the present invention;

[0034] Figure 2 is a global branch network schematic diagram of the present invention;

[0035] Figure 3 It is a schematic diagram of the regional division of pedestrian images by the local medium-granularity feature branch;

[0036] Figure 4 It is a schematic diagram of the local medium-grained characteristic branch of the present invention;

[0037] Figure 5 It is a schematic diagram of the local fine-grained graph convolution branch of the present invention. DETAILED DESCRIPTION

[0038] The present invention is further described below in conjunction with embodiments, which are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, other embodiments obtained by ordinary technicians in the field without creative work are all within the scope of protection of the present invention.

[0039] In the description of the present invention, it should be noted that the terms "center", "upper", "lower", "left", "right", "vertical", "horizontal", "inside", "outside" and the like indicate positions or positional relationships based on the positions or positional relationships shown in the drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as limiting the present invention; the terms "first", "second", and "third" are only used for descriptive purposes, and cannot be understood as indicating or implying relative importance; in addition, unless otherwise clearly specified and limited, the terms "installed", "connected", and "connected" should be understood in a broad sense, for example, it can be a fixed connection, a detachable connection, or an integral connection; it can be directly connected or indirectly connected through an intermediate medium. For those of ordinary skill in the art, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0040] The multi-granularity feature pedestrian re-identification method based on posture guidance of the present invention comprises:

[0041] (1) constructing a multi-granularity feature extraction network model, wherein the multi-granularity feature extraction network model includes a backbone feature extraction network, a human key point extractor, a global branch network (Global Branch), a local medium-granularity feature branch (Part-based local Branch) and a local fine-grained graph convolution branch (Joint-based local Branch);

[0042] (2) training the multi-granularity feature extraction network model to obtain a trained multi-granularity feature extraction network model;

[0043] (3) The image of the pedestrian to be identified is input into the trained multi-granularity feature extraction network model, and the global features, local medium-granularity features and local fine-granularity features are integrated to perform multi-granularity pedestrian re-identification.

[0044] In some embodiments, the pedestrian image to be identified is preprocessed before being input into the trained multi-granularity feature extraction network model in step (3), and the preprocessing includes using four data enhancement methods: random erasing, random padding, random cropping, and horizontal flipping, so as to improve the robustness of the model to occluded pedestrian images and prevent overfitting.

[0045] In some embodiments, the backbone feature extraction network is composed of a multi-layer convolutional neural network, which is responsible for extracting backbone network features from input pedestrian images during training and testing; the human key point extractor is composed of a multi-layer convolutional neural network, which is responsible for extracting key point information from input pedestrian images during training and testing, wherein the key point information includes key point coordinates, key point confidence, and key point heat map.

[0046] In some embodiments, the global branch network combines the backbone network features extracted by the backbone feature extraction network with the corresponding key point heat map as input and performs Hadamard product to obtain global features, and then outputs the global features as coarse-grained features through a global average pooling layer and a fully connected classification layer.

[0047] In some embodiments, the local medium-granularity feature branch uses the key point coordinates extracted by the human key point extractor and divides the body into several regions according to the actual posture of the human body, and then extracts several body regions in the pedestrian image according to the key point coordinates, and then passes the several body regions through the global average pooling layer and the fully connected classification layer as medium-granularity feature outputs.

[0048] In some embodiments, the local medium-granularity feature branch is divided into three regions according to the key point coordinates and the actual posture of the human body, and the head region, the upper body region and the lower body region are extracted, and the head region and the upper body region overlap with each other, and the upper body region and the lower body region overlap with each other;

[0049] In some embodiments, the local medium-granularity feature branch generates a corresponding bounding box B according to the key point position of each body region (head region, upper body region and lower body region). j ; where j is 1, 2, 3, representing top, middle, and bottom respectively; bounding box B j The coordinates of the upper left corner and lower right corner are calculated according to the key point position as follows:

[0050] B j =[min(x j)-5,min(y j )-5,max(x j ),max(y j )];

[0051] Among them, x j and j Represents the horizontal and vertical coordinates of the key point of the jth region.

[0052] In some embodiments, the local medium-granularity feature branch also uses the key point confidence S i , and score the visibility of each body region; for the i-th body region, its visibility score is calculated according to the following formula:

[0053]

[0054] Among them, S i is the confidence of the key points in the body area, and N is the number of key points.

[0055] In some embodiments, the local medium-granularity feature branch processes each pedestrian image to be identified and the candidate image, identifies the unobstructed shared feature area in the two images, and calculates the distance between the features based on the unobstructed shared feature area. The calculation formula is as follows:

[0056]

[0057] Where D is the cosine distance function, represents the alignment features of the i-th part of the query set image, and accordingly, represents the i-th part of the candidate set image, and dist represents the final total distance.

[0058] In some embodiments, the local fine-grained graph convolution branch combines the backbone network features extracted by the backbone feature extraction network with the corresponding key point heat map to obtain a number of key point features, and inputs the combined key point features as graph nodes, and then enters the multi-layer graph convolution network to obtain a number of learned key point features; then the learned key point features are output as fine-grained features through the global average pooling layer and the fully connected classification layer.

[0059] Combined with Figure 1 To Attachment Figure 5 , the multi-granularity feature pedestrian re-identification method based on posture guidance of the present invention is further described, and the present invention includes:

[0060] Step 1: Build a multi-granularity feature extraction network model.

[0061] like Figure 1As shown in the figure, the multi-granularity feature extraction network model consists of five parts: backbone feature extraction network, human key point extractor, global branch network (Global Branch), local medium-granularity feature branch (Part-based local Branch) and local fine-grained graph convolution branch network (Joint-based local Branch).

[0062] The model was implemented using the pytorch deep learning platform and trained using an NVIDIA RTX1080Ti. The training images were resized to 256×128 and augmented with random horizontal flipping and 10-pixel padding. Random cropping and random erasing were performed. When testing on occluded / partial datasets, we used additional color jitter augmentation to avoid domain variance. The batch size was set to 64, with 4 images per person. In the training phase, all three branches were jointly trained in an end-to-end manner for 120 epochs, with an initial learning rate of 3.5e-4, decaying to 0.1 at 30 and 70 epochs.

[0063] like Figure 2 As shown, the global branch network (Global Branch) of this embodiment combines the backbone network features with the key point heat map obtained by the human key point extractor (such as the HRNet network). Specifically, the 17 key point heat maps obtained by the human key point extractor (such as the HRNet network) are converged on a heat map to obtain the key point heat map, and the Hadamard product is performed with the backbone network features to obtain the global feature, and this feature (i.e. the global feature obtained by the Hadamard product) is globally averaged and pooled and input into the fully connected layer for classification. The specific calculation steps are as follows:

[0064]

[0065] Among them, F g is the fused global feature, m cnn is the feature extracted by the backbone network, m kp is the key point heat map, Represents the Hadamard product (element-wise multiplication).

[0066] like Figure 3 As shown in FIG. 1 , the local medium-granularity feature branch (Part-based local Branch) of this embodiment focuses on three main body regions: the head region, the upper body region, and the lower body region. Figure 3 As shown in Figure 1, the head region extends from the head to the chest, including key points such as the head, left shoulder, and right shoulder. The upper body region extends from the neck to the hips, mainly including the torso and upper limbs. The lower body region extends from the hips to the feet. In order to mitigate the impact of pose estimation errors, the three regions are designed to overlap with each other.

[0067] Then, based on the key position of each area, the corresponding bounding box B is generated j , where j can be 1, 2, or 3, representing top, middle, and bottom, respectively. Bounding box B j The coordinates of the upper left corner and lower right corner are calculated according to the key point position as follows:

[0068] B j =[min(x j )-5,min(y j )-5,max(x j ),max(y j )];

[0069] Among them, x j and j Represents the horizontal and vertical coordinates of the key point of the jth region.

[0070] Considering the diversity of pedestrian postures in real scenes, this paper introduces key point confidence scores to evaluate regional significance in the case of occlusion, thereby reducing the impact of occlusion on feature learning. For the i-th region, its visibility score P i The calculation is as follows: The visibility score of each area can be expressed as:

[0071]

[0072] Among them, S i is the confidence of the key points in the area, N is the number of key points, and N is 17 in the COCO dataset (i.e., the human key extractor proposes 17 key points). i It reflects the visibility of the area. The higher the score, the more significant the area is and the less occlusion it has. Conversely, the lower the score, the more occlusion it has. In the process of person re-identification, the weight of each area's features can be dynamically adjusted according to its significance score, thereby improving the robustness of recognition.

[0073] In order to solve the occlusion and spatial misalignment problems in feature matching methods, the local medium-granularity feature branch of this example adopts a feature alignment strategy based on visibility, focusing on processing the shared visible area between the query image and the candidate image in the feature matching stage, thereby effectively eliminating the influence of occlusion and misaligned areas. In the image matching process, only when the corresponding feature areas of the query image and the candidate image are not occluded, will the area be considered for similarity calculation and matching.

[0074] In this example, the local medium-granularity feature branch obtains the feature image representation of the query image set and the test image set respectively. Secondly, the algorithm processes each query and test image pair, identifies the unobstructed shared feature areas in the two images, and then calculates the distance between the features based only on these areas. The relevant calculation formula is as follows:

[0075]

[0076] Where D is the cosine distance function, represents the alignment features of the i-th part of the query set image, and accordingly, represents the i-th part of the candidate set image, and dist represents the final total distance.

[0077] like Figure 4 As shown in the figure, in the local medium-granularity feature branch (Part-based local Branch), a local feature extraction method based on human posture key points is used. When matching pedestrians, matching is performed based on the three-part features of the pedestrian. Since the matching is based on the position of the key points, there will be no problems of pedestrian feature matching confusion such as the head matching the upper body, the upper body matching the legs, etc., during matching. This effectively avoids the problem of horizontal block division requiring pedestrian height alignment. At the same time, the pedestrian feature map is divided into three different blocks, and there is a certain overlap between the key points of two adjacent parts, which ensures that the parts are interconnected. Different body regions have different visual features and semantic information. F represents the output feature map of the upper, middle, and lower body branches. and

[0078] like Figure 5 As shown in the figure, the local fine-grained graph convolution branch, given a pedestrian image X, extracts its feature map m through the backbone feature extraction network (ResNet50) cnn And the key point heat map m obtained by the human key point extractor (HRNet network) kp These two features represent the semantic information and key point information of the pedestrian image respectively.

[0079] In order to construct the input of the graph convolutional network, the Hadamard product And the global average pooling operation (g(·)) is used to process the feature map. Specifically, the key point map m kp And feature map m cnn Perform an external product operation to obtain a set of key point areas These local features represent the semantic information related to the key points in the pedestrian image. cnn Perform global average pooling operation to obtain global features These global features represent the semantic information of the entire pedestrian image. The above process can be expressed by the following equation:

[0080]

[0081] In this way, local and global features of pedestrian images can be obtained. These features will be used to establish connections between pedestrian nodes and transfer information in the following steps.

[0082] In order to better learn the association relationship between nodes, this example introduces a graph convolution layer, which enables it to adaptively learn the adjacent matrix (node ​​links) based on the input features.

[0083] For the graph convolution layer, the specific process is as follows:

[0084] (1) First, local features and global features The global features corrected by learning are obtained by concatenation through the fully connected layer, and then the difference features between the local features and the corrected global features are calculated. A large difference value indicates that the local feature is very different from the global feature, which may be outlier information caused by occlusion or noise; a small difference value indicates that the local feature is highly correlated with the global feature and is an important regional feature.

[0085] (2) By performing element-wise multiplication operation on the extracted difference feature matrix A and the predefined graph adjacency matrix DA, a new adjacency matrix representing the relationship between key points can be obtained. Perform matrix multiplication for further processing.

[0086] (3) Fusion of local features After the relationship features are combined, a topological graph structure feature containing rich relationship information is formed. The local features of the key points obtained after processing not only contain their own attribute information, but also integrate the feature data of adjacent nodes to obtain the final output F j .

[0087] Step 2: Train the multi-granularity feature extraction network model to obtain a trained multi-granularity feature extraction network model.

[0088] In person re-identification, the loss function plays a vital role. The loss function is used to supervise the training process, so that the model can reach a convergence state and reduce the error of the model prediction value.

[0089] This embodiment uses triple loss and label smoothing loss for joint training, and the formula is as follows:

[0090]

[0091] In the above formula, K represents the number of pedestrians, N represents the size of a batch input, and y i represents the identity label of the i-th pedestrian image, x i represents the features of the i-th image sample in a batch, w j represents the jth column of the weight matrix, b represents the bias term, represents the target sample, represents a positive sample, Represents negative samples.

[0092] The final global branch loss, local medium-grained feature branch loss, and local fine-grained graph convolution branch loss are:

[0093] L global =L id (FC(GAP(F g )))+λL tri (FC(GAP(F g )))

[0094]

[0095] Among them, GAP represents the global average pooling operation, FC represents the fully connected operation, and P i is the regional visibility score, C k is the key point confidence score.

[0096] Step 3: Input the pedestrian image to be identified into the trained multi-granularity feature extraction network model, fuse the global features, local medium-granularity features and local fine-granularity features, and then perform multi-granularity pedestrian re-identification. For the fusion of multi-granularity features, the weight calculation formula can be added:

[0097] F final =α·F global +β·F local_mid +γ·F local_fine

[0098] Among them, F final is the final fusion feature, α, β, γ are the weight coefficients of each feature, and α+β+γ=1, F global is the global feature, F local_mid is the medium-grained local feature, F local_fine It is a fine-grained local feature.

[0099] In order to more accurately and comprehensively evaluate the performance of this model, the present invention uses two commonly used evaluation indicators Rank-1 and mAP for pedestrian re-identification. Experiments have shown that the recognition effect of the present invention is higher than the evaluation indicators of most current mainstream models. On the Occluded-Duke dataset, Rank-1 and mAP reached 70.4% and 56.7% respectively, which is a very high recognition accuracy in the industry.

[0100] It should be understood that the above description of the preferred embodiment is relatively detailed and cannot be regarded as limiting the scope of patent protection of the present invention. Under the enlightenment of the present invention, ordinary technicians in this field can also make substitutions or modifications without departing from the scope of protection of the claims of the present invention, which all fall within the scope of protection of the present invention. The scope of protection requested for the present invention shall be based on the attached claims.

Claims

1. A multi-granularity feature pedestrian re-identification method based on posture guidance, characterized in that: include: (1) constructing a multi-granularity feature extraction network model, wherein the multi-granularity feature extraction network model includes a backbone feature extraction network, a human key point extractor, a global branch network, a local medium-granularity feature branch, and a local fine-grained graph convolution branch; (2) training the multi-granularity feature extraction network model to obtain a trained multi-granularity feature extraction network model; (3) The image of the pedestrian to be identified is input into the trained multi-granularity feature extraction network model, and the global features, local medium-granularity features and local fine-granularity features are integrated to perform multi-granularity pedestrian re-identification.

2. The multi-granularity feature pedestrian re-identification method based on posture guidance according to claim 1 is characterized in that: In step (3), the pedestrian image to be identified is preprocessed before being input into the trained multi-granularity feature extraction network model. The preprocessing includes using four data enhancement methods: random erasing, random padding, random cropping, and horizontal flipping, so as to improve the robustness of the model to occluded pedestrian images and prevent overfitting.

3. The method for pedestrian re-identification based on posture guidance and multi-granularity features according to claim 1 is characterized in that: The backbone feature extraction network is composed of a multi-layer convolutional neural network, which is responsible for extracting backbone network features from input pedestrian images during training and testing; the human key point extractor is composed of a multi-layer convolutional neural network, which is responsible for extracting key point information from input pedestrian images during training and testing, wherein the key point information includes key point coordinates, key point confidence, and key point heat map.

4. The method for pedestrian re-identification based on posture guidance and multi-granularity features according to claim 3 is characterized in that: The global branch network combines the backbone network features extracted by the backbone feature extraction network with the corresponding key point heat map as input and performs Hadamard product to obtain the global features. The global features are then output as coarse-grained features through the global average pooling layer and the fully connected classification layer.

5. The method for pedestrian re-identification based on posture guidance and multi-granularity features according to claim 3 is characterized in that: The local medium-granularity feature branch uses the key point coordinates extracted by the human key point extractor and divides the human body into several regions according to the actual posture, and then extracts several body regions in the pedestrian image according to the key point coordinates. The several body regions are then output as medium-granularity features after passing through a global average pooling layer and a fully connected classification layer.

6. The method for pedestrian re-identification based on posture guidance and multi-granularity features according to claim 5 is characterized in that: The local medium-granularity feature branch is divided into three regions according to the key point coordinates and the actual posture of the human body, and the head region, the upper body region and the lower body region are extracted, and the head region overlaps with the upper body region, and the upper body region overlaps with the lower body region.

7. The method for pedestrian re-identification based on posture guidance and multi-granularity features according to claim 6 is characterized in that: The local medium-granularity feature branch generates the corresponding bounding box B according to the key point position of each body region. j ; where j is 1, 2, 3, representing top, middle, and bottom respectively; bounding box B j The coordinates of the upper left corner and lower right corner are calculated according to the key point position as follows: B j =[min(x j )-5,min(y j )-5,max(x j ),max(y j )]; Among them, x j and j Represents the horizontal and vertical coordinates of the key point of the jth region.

8. The method for pedestrian re-identification based on posture guidance and multi-granularity features according to claim 7 is characterized in that: The local medium-granularity feature branch also uses the key point confidence S i , and score the visibility of each body region; for the i-th body region, its visibility score is calculated according to the following formula: Among them, S i is the confidence of the key points in the body area, and N is the number of key points.

9. The method for pedestrian re-identification based on posture guidance and multi-granularity features according to claim 8 is characterized in that: The local medium-granularity feature branch processes each pedestrian image to be identified and the candidate image, identifies the unobstructed shared feature area in the two images, and calculates the distance between the features based on the unobstructed shared feature area. The calculation formula is as follows: Where D is the cosine distance function, represents the alignment features of the i-th part of the query set image, and accordingly, represents the i-th part of the candidate set image, and dist represents the final total distance.

10. The method for pedestrian re-identification based on posture guidance and multi-granularity features according to claim 3 is characterized in that: The local fine-grained graph convolution branch combines the backbone network features extracted by the backbone feature extraction network with the corresponding key point heat map to obtain several key point features, and inputs the combined key point features as graph nodes, and then enters the multi-layer graph convolution network to obtain several learned key point features; then the learned key point features are output as fine-grained features through the global average pooling layer and the fully connected classification layer.

Images