Target re-identification method, device, equipment and computer-readable storage medium

Abstract

The present application is applicable to the field of image recognition technology, and provides a target re-identification method, device, equipment and computer-readable storage medium, the method comprising: obtaining multiple frames of images, each of which includes a target to be identified; inputting a single frame of the multiple frames into an appearance feature extraction network to obtain the appearance features of the target to be identified, the appearance features of the target to be identified are independent of the color of the single frame, and the appearance feature extraction network is used to eliminate the interference of the color of the image on the appearance of the target; according to the appearance features of the target to be identified, determine an image that matches the target to be identified from a historical image library. This method can avoid the problem of low re-identification accuracy of the target image to be identified due to the different clothing colors of the target image to be identified and the target image in the historical image library.

Description

Technical Field

[0001] The present application belongs to the field of image recognition technology, and in particular, relates to a target re-identification method, device, equipment and computer-readable storage medium. Background Art

[0002] Target re-identification technology is an important technical means in the fields of intelligent security, missing target search, case investigation, etc. For example, in the field of missing target search, it is used to find the missing target based on the image of the missing target.

[0003] At present, the relevant technologies mainly use deep learning technology for target re-identification. Specifically, the Yolo detection algorithm is first used to detect each image in the multiple frames included in multiple video files and crop and save them to form a historical image library; then the feature extraction network is used to extract the feature vector of each image in the cropped historical image library; finally, the image in the historical image library is compared to find the image whose feature vector is closest to the target image to be identified.

[0004] However, since the clothing colors of the target image to be identified and the target image in the historical image library may be different, the recognition accuracy of the target image to be identified in the related technology is low. Summary of the invention

[0005] The present application provides a target re-identification method, apparatus, device and computer-readable storage medium, which can avoid the problem of low re-identification accuracy of a target image to be identified due to the difference in clothing color between the target image to be identified and the target image in the historical image library.

[0006] In a first aspect, the present application provides a target re-identification method, comprising:

[0007] Acquire multiple frames of images, each frame of the image includes the target to be identified;

[0008] Inputting a single frame image from the multiple frames into an appearance feature extraction network to obtain the appearance features of the target to be identified, wherein the appearance features of the target to be identified are independent of the color of the single frame image, and the appearance feature extraction network is used to eliminate the interference of the color of the image on the appearance of the target;

[0009] According to the appearance features of the target to be identified, an image matching the target to be identified is determined from a historical image library.

[0010] The present application obtains the appearance features of the target to be identified that are independent of the color of the single-frame image by inputting a single-frame image from a multi-frame image into an appearance feature extraction network, and determines an image matching the target to be identified from a historical image library based on the appearance features of the target to be identified. This avoids the problem of low re-identification accuracy of the target image to be identified due to the different clothing colors of the target image to be identified and the target image in the historical image library, and improves the re-identification accuracy.

[0011] In a second aspect, the present application provides a target re-identification device, which is used to execute the method in the first aspect or any possible implementation of the first aspect. Specifically, the device includes:

[0012] A first acquisition module is used to acquire multiple frames of images, each of which includes a target to be identified;

[0013] A second acquisition module is used to input a single frame image of the multiple frames into an appearance feature extraction network to obtain the appearance features of the target to be identified, wherein the appearance features of the target to be identified are independent of the color of the single frame image, and the appearance feature extraction network is used to eliminate the interference of the color of the image on the appearance of the target;

[0014] The recognition module is used to determine an image matching the target to be recognized from a historical image library according to the appearance features of the target to be recognized.

[0015] In a third aspect, the present application provides a target re-identification device, the device comprising a memory and a processor. The memory is used to store instructions; the processor executes the instructions stored in the memory, so that the device executes the target re-identification method in the first aspect or any possible implementation of the first aspect.

[0016] In a fourth aspect, a computer-readable storage medium is provided, wherein instructions are stored in the computer-readable storage medium. When the instructions are executed on a computer, the computer executes the target re-identification method in the first aspect or any possible implementation of the first aspect.

[0017] In a fifth aspect, a computer program product comprising instructions is provided. When the instructions are executed on a device, the device executes the target re-identification method in the first aspect or any possible implementation manner of the first aspect.

[0018] It can be understood that the beneficial effects of the second to fifth aspects mentioned above can be found in the relevant description of the first aspect mentioned above, and will not be repeated here. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] In order to more clearly illustrate the technical solution in the present application, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0020] Figure 1 is a flowchart of a target re-identification method provided by an embodiment of the present application;

[0021] Figure 2 is a flowchart of a target re-identification method provided by an embodiment of the present application;

[0022] Figure 3 is a flowchart of a target re-identification method provided by an embodiment of the present application;

[0023] Figure 4 is a flowchart of a target re-identification method provided by an embodiment of the present application;

[0024] Figure 5a is a flowchart of a target re-identification method provided by an embodiment of the present application;

[0025] Figure 5b is a schematic diagram of determining a first score provided by an embodiment of the present application;

[0026] Figure 5c is a schematic diagram of determining a second score provided by an embodiment of the present application;

[0027] Figure 6 is a flowchart of a target re-identification method provided by an embodiment of the present application;

[0028] Figure 7 is a schematic diagram of the structure of a target re-identification device provided in an embodiment of the present application;

[0029] Figure 8 It is a structural diagram of a target re-identification device provided in one embodiment of the present application. DETAILED DESCRIPTION

[0030] In the following description, specific details such as specific system structures and technologies are provided for the purpose of illustration rather than limitation, so as to provide a thorough understanding of the present application. However, it should be clear to those skilled in the art that the present application may 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 prevent unnecessary details from obstructing the description of the present application.

[0031] It should be understood that when used in the present specification and the appended claims, the term "comprising" indicates the presence of described 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, components and / or combinations thereof.

[0032] It should also be understood that the term “and / or” used in the specification and appended claims refers to any and all possible combinations of one or more of the associated listed items, and includes these combinations.

[0033] As used in the specification and appended claims of this application, the term "if" can be interpreted as "when" or "uponce" or "in response to determining" or "in response to detecting", depending on the context. Similarly, the phrase "if it is determined" or "if [described condition or event] is detected" can be interpreted as meaning "uponce it is determined" or "in response to determining" or "uponce [described condition or event] is detected" or "in response to detecting [described condition or event]", depending on the context.

[0034] In addition, in the description of the present application specification and the appended claims, the terms "first", "second", "third", etc. are only used to distinguish the descriptions and cannot be understood as indicating or implying relative importance.

[0035] References to "one embodiment" or "some embodiments" etc. described in the specification of this application mean that one or more embodiments of the present application include specific features, structures or characteristics described in conjunction with the embodiment. Therefore, the statements "in one embodiment", "in some embodiments", "in some other embodiments", "in some other embodiments", etc. that appear in different places in this specification do not necessarily refer to the same embodiment, but mean "one or more but not all embodiments", unless otherwise specifically emphasized in other ways. The terms "including", "comprising", "having" and their variations all mean "including but not limited to", unless otherwise specifically emphasized in other ways.

[0036] The present application provides a target re-identification method, device, equipment and computer-readable storage medium. The method can be implemented by a target re-identification device and applied to scenarios such as case investigation, missing person search and intelligent security.

[0037] The target re-identification device is connected to the target re-identification device in communication. For example, the target re-identification device can communicate with the target re-identification device through an application (APP), a web page, a public account in the application, and a mini-program, so that the target re-identification device and the target re-identification device can transmit information to each other. The user can achieve target re-identification through the target re-identification device that communicates with the target re-identification device.

[0038] Among them, the target re-identification device refers to the device used by the user to perform target re-identification. The target re-identification device can be a device with display hardware and corresponding software support, such as a smart phone, a tablet computer, a desktop computer, a laptop computer, a wearable device, a handheld device, and a vehicle-mounted device. The embodiment of the present application does not impose any restrictions on the specific type of the target re-identification device.

[0039] Based on the above scenario description, the target re-identification method provided in the embodiment of the present application is described in detail below in combination with the target re-identification device.

[0040] See also Figure 1 , Figure 1 A schematic flow chart of a target re-identification method provided in an embodiment of the present application is shown.

[0041] like Figure 1 As shown, the target re-identification method provided by the present application may include:

[0042] S101, acquiring multiple frames of images, each frame of the image including a target to be identified.

[0043] In some embodiments, the target re-identification device may display a search page of the target re-identification device. The search page of the target re-identification device is used to display an input port for inputting multiple frames of images and to display search results. The present application does not limit the specific implementation of the search page of the target re-identification device. In some embodiments, the search page may include a control for inputting a video or an image, and the control is used to trigger a search by inputting a video or an image.

[0044] The user can perform target re-identification in the search page of the target re-identification device. Thus, the target re-identification device can send the user's search request to the target re-identification device. The user's search request is used to indicate that the user wants to perform target re-identification.

[0045] For example, after receiving an operation such as a click, double-click, or long press on a control for inputting a video or image as instructed by the user, the target re-identification device may convert the operation into a search request of the user and send the search request of the user to the target re-identification device. This application does not limit the specific implementation method of the user's search request.

[0046] The multiple frames of images are obtained by tracking the target to be identified in given video data through a target tracking algorithm.

[0047] The target re-identification device may store the target tracking algorithm in the target re-identification device and / or a storage device.

[0048] The storage device can communicate with the target re-identification apparatus, so that the target re-identification apparatus can obtain the appearance features extracted by the appearance feature extraction network from the storage device. This application does not limit the storage method and specific type of the storage device.

[0049] In some embodiments, the target tracking algorithm is the DeepSORT algorithm.

[0050] The given video data may be directly given by a user, or may be extracted from video data collected by image acquisition devices such as surveillance cameras and video cameras.

[0051] The camera may include a single camera, a dual camera, or a triple camera, or the camera may be set as a wide-angle camera or a telephoto camera, which is not limited in the embodiments of the present application.

[0052] In some embodiments, the multiple frames of images are images from video data directly given by a user. Specifically, in the given multiple frames of images, each frame of the image includes a target to be identified.

[0053] In other embodiments, the multiple frames of images are extracted from video data collected by image acquisition devices such as surveillance cameras and video cameras. Specifically, each of the multiple frames of images collected includes a target to be identified.

[0054] In some embodiments, when the identity of the target to be identified is known, each frame of the multiple frames of images includes a target to be identified whose identity is known.

[0055] In some embodiments, when the identity of the target to be identified is unknown, each frame of the multiple frames of images includes a target to be identified with an unknown identity.

[0056] The target to be identified includes but is not limited to the human body.

[0057] In a specific embodiment, assuming that the target re-identification device is a mobile phone, and the target re-identification device is a mini program, the mini program can display a search page for the target re-identification device. In the missing person search scenario, the user can input multiple frames of images, each of which includes the missing person A to be identified, in the search page of the mini program.

[0058] S102. Input a single-frame image from the multiple-frame images into an appearance feature extraction network to obtain appearance features of the target to be identified. The appearance features of the target to be identified are independent of the color of the single-frame image. The appearance feature extraction network is used to eliminate interference of the color of the image on the appearance of the target.

[0059] Based on S101, the target re-identification device may obtain a search request. Thus, the target re-identification device may execute the search request by first inputting a single frame image from the multiple frames of images into an appearance feature extraction network to extract the appearance features of the target to be identified.

[0060] The appearance feature extraction network is pre-designed and is used to eliminate the interference of the color of the image on the appearance of the target. The appearance feature extraction network can be insensitive to the color of clothing when extracting appearance features.

[0061] The target re-identification device may store the pre-designed appearance feature extraction network in the target re-identification device and / or a storage device.

[0062] In some embodiments, the appearance features may include, but are not limited to, one or more of hair, tops, bottoms, gender, whether a backpack is worn, whether a hat is worn, whether a backpack is worn, and shoes. For example, the appearance features include hair, tops, bottoms, gender, and shoes.

[0063] In a specific embodiment, assuming that the target re-identification device is a mobile phone and the target re-identification device is a mini program, the mini program can display a search page for the target re-identification device. In the missing person search scenario, after the user enters multiple frames of images in which each frame includes the missing person A to be identified in the mini program search page, the mini program retrieves the appearance feature extraction network from the storage device to extract the hair, top, bottom, gender and shoes features of the missing person A.

[0064] S103: Determine, from a historical image library, an image matching the target to be identified according to the appearance features of the target to be identified.

[0065] The images in the historical image library can be an image library with multiple historical target images pre-stored in the target re-identification device / storage device, or they can be images corresponding to video data collected from multiple image acquisition devices such as surveillance cameras and video cameras that are communicatively connected to the target re-identification device.

[0066] In some embodiments, when the identity of the target to be identified is known, an image matching the target to be identified may be matched from a pre-stored image library having a plurality of historical target images.

[0067] In other embodiments, when the identity of the target to be identified is unknown, an image matching the target to be identified can be matched from images corresponding to video data captured by multiple image acquisition devices such as surveillance cameras and video cameras that are communicatively connected to the target re-identification device.

[0068] In a specific embodiment, assuming that the target re-identification device is a mobile phone and the target re-identification apparatus is a mini program, the mini program can display a search page for the target re-identification apparatus. In the missing person search scenario, after the user enters multiple frames of images in which each frame includes the missing person A to be identified in the mini program search page, the mini program retrieves the appearance feature extraction network from the storage device to extract the hair, top, bottom, gender and shoe features of the missing person A, retrieves the historical image library from the storage device, and searches for images matching the missing person A from the historical image library based on the hair, top, bottom, gender and shoe features of the missing person A, and displays the images matching the missing person A found from the historical image library on the search page.

[0069] The target re-identification method provided by the present application obtains the appearance features of the target to be identified that are independent of the color of the single-frame image by inputting a single-frame image from a multi-frame image into an appearance feature extraction network, and determines an image matching the target to be identified from a historical image library based on the appearance features of the target to be identified. The appearance features of the target to be identified that are independent of color in a single-frame image are extracted by means of an appearance feature extraction network that is used to eliminate the interference of image color on the target's appearance. Since the appearance features of the target to be identified are independent of color, the problem of low re-identification accuracy of the target image to be identified due to the different clothing colors of the target image to be identified and the target image in the historical image library can be avoided.

[0070] Based on the above Figure 1 In addition to the description of the illustrated embodiment, the present application also provides a generation process of a pre-designed appearance feature extraction network.

[0071] Next, combine Figure 2 , which introduces in detail the specific implementation process of the process of generating the appearance feature extraction network of this application.

[0072] based on Figure 1 As described in S102, when the target re-identification device obtains the appearance features in a single-frame image, the appearance feature extraction network is used to eliminate the interference of the color of the image on the appearance of the target.

[0073] The appearance feature extraction network generation process can be generated by a feature network generation system or by other feasible network generation systems, which will not be described in detail here.

[0074] See also Figure 2 , Figure 2A schematic flow chart of a process for generating an appearance feature extraction network provided in an embodiment of the present application is shown.

[0075] like Figure 2 As shown, the process of generating an appearance feature extraction network provided by the present application may include:

[0076] S201. Acquire at least one frame of sample image, each frame of the sample image including a sample target.

[0077] In some embodiments, the sample image is a plurality of frames.

[0078] The sample image may be an image prepared in advance, or may be selected from existing images each frame of which includes a sample target.

[0079] S202: Perform data enhancement on each pixel in each frame of the sample image to obtain multiple frames of enhanced images.

[0080] In some embodiments, multiple frames of enhanced images are obtained by exchanging color values ​​of each color channel of each pixel in each frame of the sample image.

[0081] For example, when the sample image is an RGB image, the color values ​​of the RGB color channels of each pixel in the RGB image are exchanged to obtain five images except the original RGB image, including: an RBG image, a GRB image, a GBR image, a BRG image and a BGR image.

[0082] In other embodiments, multiple frames of enhanced images are obtained by exchanging the color values ​​of each color channel for each pixel in each frame of the sample image, and converting the grayscale value of each pixel in each frame of the sample image.

[0083] The grayscale value of each pixel in each frame of the sample image is converted, that is, the grayscale value of each pixel is randomly changed.

[0084] For example, each pixel with a grayscale value of 10 in the sample image is converted into a pixel with a grayscale value of 20 or 8, and each pixel with a grayscale value of 40 in the sample image is converted into a pixel with a grayscale value of 50 or 45.

[0085] S203: Determine appearance features of the sample object according to the multiple frames of enhanced images, where the appearance features of the sample object are independent of colors of the multiple frames of enhanced images.

[0086] S204: training an original feature extraction network according to the appearance features of the sample target to obtain the appearance feature extraction network.

[0087] In the present application, in the process of generating the appearance extraction network by the feature network generation system, a sample image is first obtained, and data enhancement is performed on each pixel point in each frame of the sample image to obtain a multi-frame enhanced image. Based on the multi-frame enhanced image, the appearance features of the sample target are determined, and the appearance features of the sample target are independent of the color of the multi-frame enhanced image. Based on the appearance features of the sample target, the original feature extraction network is trained to obtain an appearance feature extraction network. Since data enhancement processing is performed on each pixel point of the sample image during the generation process, the network trained by this process can be used to eliminate the interference of the color of the image on the appearance of the target.

[0088] based on Figure 1 As described in the illustrated embodiment, the target re-identification device may also introduce gait features when implementing target re-identification, and combine the gait features with the appearance features to determine an image that matches the target to be identified from the historical image library.

[0089] Among them, gait features are designed to identify people through their posture and movements when walking. Gait recognition is not easy to disguise and is not sensitive to clothing.

[0090] Next, combine Figure 3 , the specific implementation method of the target re-identification device to perform the above process is introduced in detail.

[0091] See also Figure 3 , Figure 3 A schematic flow chart of a target re-identification method provided in an embodiment of the present application is shown.

[0092] like Figure 3 As shown, the target re-identification method provided by the present application may include:

[0093] S301, acquiring multiple frames of images, each frame of image including a target to be identified.

[0094] S302. Input a single-frame image from the multiple-frame images into an appearance feature extraction network to obtain appearance features of the target to be identified. The appearance features of the target to be identified are independent of the color of the single-frame image. The appearance feature extraction network is used to eliminate interference of the color of the image on the appearance of the target.

[0095] Among them, S301 and S302 are respectively Figure 1 The implementation methods of S101 and S102 in the illustrated embodiment are similar and will not be described in detail in this application.

[0096] S303: Acquire temporally continuous frame images from the multiple frame images.

[0097] The temporally continuous frame images represent images that are continuous in time sequence.

[0098] In some embodiments, continuous frame images that are continuous in time are obtained by intercepting images that are continuous in time sequence from multiple frame images.

[0099] In a specific embodiment, assuming that the target re-identification device is a mobile phone and the target re-identification device is a mini program, the mini program can display a search page for the target re-identification device. In the missing person search scenario, after the user enters multiple frames of images in which each frame includes the missing person A to be identified in the search page of the mini program, the mini program captures continuous frame images from the multiple frames of images of the missing person A to be identified.

[0100] S304: Determine the gait features of the target to be identified based on the continuous frame images, where the gait features of the target to be identified are used to represent the posture and movement of the target to be identified when walking.

[0101] In some embodiments, the target re-identification device extracts the gait features of the target to be identified from a gait feature extraction network based on continuous frame images.

[0102] The target re-identification device stores the gait feature extraction network in the target re-identification device and / or a storage device.

[0103] In some implementations, by inputting multiple frames of images into a gait feature extraction network, the gait features of the target to be identified can be obtained.

[0104] In some embodiments, the formula for extracting appearance features through the gait feature extraction network is:

[0105] f i =F(X i )

[0106]

[0107] Among them, f i represents the continuous frame image X of the target identity i obtained by the conversion function of the gait feature extraction network i gait features, F represents the conversion function of the gait feature extraction network, X i represents the continuous frame image of the target to be identified as i, n represents the total number of continuous frame images, P represents the identity of the target to be identified, P i represents the target to be identified with identity i, represents the jth frame image of the target to be identified with identity i in the continuous frame images, Indicates that, P i The j-th frame image.

[0108] In a specific embodiment, assuming that the target re-identification device is a mobile phone and the target re-identification apparatus is a mini program, the mini program can display a search page for the target re-identification apparatus. In the missing person search scenario, after the user enters multiple frames of images in which each frame includes the missing person A to be identified in the mini program search page, the mini program calls the gait feature extraction network from the storage device to extract the walking motion and posture features of the missing person A from the continuous frame images.

[0109] S305 . Determine, from the historical image library, an image that matches the target to be identified based on the appearance features of the target to be identified and the gait features of the target to be identified.

[0110] In a specific embodiment, assuming that the target re-identification device is a mobile phone and the target re-identification device is a mini program, the mini program can display the search page of the target re-identification device. In the missing person search scenario, after the user enters multiple frames of images in which each frame includes the missing person A to be identified in the mini program search page, the mini program retrieves the appearance feature extraction network from the storage device to extract the hair, top, bottom, gender and shoe features of each frame of the multiple frames of images including the missing person A. At the same time, the mini program retrieves the gait feature extraction network from the storage device to extract the walking motion and posture features of the missing person A from the continuous frame images. The mini program searches for human images matching the missing person A from the historical image library based on the hair, top, bottom, gender and shoe features of the missing person A and the walking motion and posture features, and displays the human images matching the missing person A found from the historical image library on the search page.

[0111] In the present application, the target re-identification device inputs a single-frame image from a multi-frame image into an appearance feature extraction network to obtain the appearance features of the target to be identified that are independent of the color of the single-frame image, determines the gait features of the target to be identified based on the appearance features of the target to be identified and through the continuous frame images, and then determines the image that matches the target to be identified from the historical image library based on the appearance features and gait features of the target to be identified. Combining the appearance features and gait features to determine the image that matches the target to be identified from the historical image library improves the efficiency of identification and the accuracy of identification; at the same time, because gait recognition has the characteristics of not being easy to disguise and not being sensitive to clothing, it can also avoid the problem of low accuracy of re-identification of the target image to be identified due to the different clothing colors of the target image to be identified and the target image in the historical image library.

[0112] based on Figure 3 According to the description of the illustrated embodiment, the target re-identification device may also obtain the gait features of the target to be identified from the continuous frame images after obtaining the continuous frame images that are continuous in time from the multiple frame images.

[0113] Next, combine Figure 4 , the specific implementation method of the target re-identification device to perform the above process is introduced in detail.

[0114] See also Figure 4 , Figure 4 A schematic flow chart of a target re-identification method provided in an embodiment of the present application is shown.

[0115] like Figure 4 As shown, the target re-identification method provided by the present application may include:

[0116] S401, separating a foreground and background region and a background region of each frame image in the continuous frame images, wherein the foreground and background region includes the target to be identified, and the background region does not include the target to be identified.

[0117] The foreground and background refer to the area that sets off the subject in the image, and serve the subject; the background refers to the area behind the subject that serves as a foil to the subject, and has the function of setting off the subject. For example, in an image of a person standing on the beach, the area including the person is the foreground, and the area of ​​the sea behind the person that does not include the person is the background.

[0118] Separating the foreground area from the background area refers to separating the area including the target to be identified from the area not including the target to be identified.

[0119] In a specific embodiment, assuming that the target re-identification device is a mobile phone and the target re-identification device is a mini program, the mini program can display a search page for the target re-identification device. In the missing person search scenario, after the user enters multiple frames of images in which each frame includes the missing person A to be identified in the mini program search page, the mini program captures continuous frame images from the multiple frames, and then separates the area including the missing person A from the area not including the missing person A in the continuous frame images.

[0120] S402: Use the foreground and background area of ​​each frame image in the continuous frame images as the continuous frame foreground and background image.

[0121] In a specific embodiment, assuming that the target re-identification device is a mobile phone, and the target re-identification device is a mini program, the mini program can display a search page for the target re-identification device. In the missing person search scenario, after the user enters multiple frames of images in which each frame includes the missing person A to be identified in the mini program search page, the mini program captures continuous frame images from the multiple frames, separates the area including the missing person A from the area not including the missing person A in the continuous frame images, and uses the area including the missing person A as the background image for the continuous frames.

[0122] S403: Determine the gait features of the target to be identified according to the background images before the continuous frames.

[0123] In a specific embodiment, assuming that the target re-identification device is a mobile phone and the target re-identification device is a mini program, the mini program can display a search page for the target re-identification device. In the missing person search scenario, after the user enters multiple frames of images in which each frame includes the missing person A to be identified in the mini program search page, the mini program captures continuous frame images from the multiple frames, separates the area including the missing person A from the area not including the missing person A in the continuous frame images, and uses the area including the missing person A as the continuous frame foreground background image. The mini program retrieves the gait feature extraction network from the storage device to extract the walking movement and posture features of the missing person A from the continuous frame foreground background image.

[0124] In the present application, the target re-identification device separates the foreground and background regions of each frame image in the continuous frame image and the background region, and uses the foreground and background regions of each frame image in the continuous frame image as the continuous frame foreground and background image, and determines the gait features of the target to be identified according to the continuous frame foreground and background image. The continuous frame foreground and background images are extracted from the continuous frame images, and the gait features are then extracted from the continuous frame foreground and background images, thereby avoiding interference from the background region and facilitating rapid acquisition of the gait features of the target to be identified.

[0125] based on Figure 4 In the description of the illustrated embodiment, when the target re-identification device combines gait features with appearance features to determine an image matching the target to be identified from a historical image library, it can do so in a variety of ways, such as introducing similarity for matching.

[0126] Next, combine Figure 5a , the specific implementation method of the target re-identification device to perform the above process is introduced in detail.

[0127] Based on the description of S304, the target re-identification device can introduce similarity, and determine the image matching the target to be identified from the historical image library by combining the similarity between the gait features of the target to be identified and the gait features of each frame image in the historical image library, as well as the similarity between the appearance features of the target to be identified and the appearance features of each frame image in the historical image library.

[0128] Here, similarity refers to the degree of similarity between two targets.

[0129] See also Figure 5a , Figure 5a A schematic flow chart of a target re-identification method provided in an embodiment of the present application is shown.

[0130] like Figure 5a As shown, the target re-identification method provided by the present application may include:

[0131] S501, acquiring multiple frames of images, each frame of image including a target to be identified.

[0132] S502. Input a single-frame image from the multiple-frame images into an appearance feature extraction network to obtain appearance features of the target to be identified. The appearance features of the target to be identified are independent of the color of the single-frame image. The appearance feature extraction network is used to eliminate interference of the color of the image on the appearance of the target.

[0133] S503: Acquire temporally continuous frame images from the multiple frame images.

[0134] S504: Determine the gait features of the target to be identified according to the continuous frame images.

[0135] Among them, S501, S502, S503 and S504 are respectively Figure 3 The implementation methods of S301, S302, S303 and S304 in the illustrated embodiment are similar and will not be described in detail in this application.

[0136] S505: Determine the appearance features and gait features of the target in each frame of the image in the historical image library.

[0137] The appearance features and gait features of the target in each frame of the historical image library may be extracted in advance, or may be extracted by the target re-identification device during the target re-identification process.

[0138] In a specific embodiment, assuming that the target re-identification device is a mobile phone and the target re-identification means is a mini program, the mini program can display a search page for the target re-identification means. In the missing person search scenario, after the user enters multiple frames of images in which each frame includes the missing person A to be identified in the mini program search page, the mini program can obtain the hair, top, bottom, gender and shoe features of the human body in each frame of the image, as well as the movement and posture features when walking, from the historical image library.

[0139] S506. Determine a plurality of first scores according to the appearance features of the target to be identified and the appearance features of the target in each frame of the image, each first score being used to indicate a similarity between the appearance features of the target to be identified and the appearance features of the target in each frame of the image.

[0140] Each first score represents the similarity between the appearance features of the target to be identified and the appearance features of the target in each frame of the image.

[0141] In some embodiments, the similarity between the appearance features of the target to be identified and the appearance features of the target in each frame of image refers to the cosine similarity between the appearance features of the target to be identified and the appearance features of the target in each frame of image.

[0142] Among them, cosine similarity, also known as cosine similarity, is the similarity obtained by calculating the cosine value of the angle between the vectors corresponding to two features.

[0143] Cosine similarity is the product of the vectors corresponding to the two features multiplied by the product of the two vector norms. The larger the cosine similarity, the smaller the angle between the vectors corresponding to the two features, and the closer the vectors corresponding to the two features are; otherwise, the farther they are.

[0144] In some embodiments, the cosine similarity calculation formula between the appearance features of the target to be identified and the appearance features of the target in each frame of image is:

[0145]

[0146] Among them, s′ represents the cosine similarity between the appearance features of the target to be identified and the appearance features of the target in each frame image, and f′ i Represents the vector corresponding to the appearance feature of the target to be identified, f′ j Represents the vector corresponding to the appearance features of the target in each frame of the image, ||f′ i ||||f′ j || represents the modulus of the vector corresponding to the appearance features of the target to be identified and the vector corresponding to the appearance features of the target in each frame image.

[0147] In a specific embodiment, assuming that the target re-identification device is a mobile phone, and the target re-identification apparatus is a mini program, the mini program can display a search page for the target re-identification apparatus. In the missing person search scenario, after the user enters multiple frames of images in which each frame includes the missing person A to be identified in the search page of the mini program, the mini program can extract the appearance features of the target in each frame from the historical image library, and calculate the cosine similarity of the hair, top, bottom, gender, and shoe features of the missing person A to be identified with the hair, top, bottom, gender, and shoe features of the human body in each frame in the historical image library.

[0148] When there are 100 frames of images in the historical image library, the cosine similarity between the hair, top, bottom, gender and shoe features of the missing person A to be identified and the hair, top, bottom, gender and shoe features of the human body in each of the 100 frames of images is calculated.

[0149] like Figure 5bAs shown, the first column is the appearance features of the missing person A to be identified, and the vector corresponding to the appearance features of the missing person A to be identified can be obtained through the appearance features of the missing person A to be identified; the second column is the appearance features of the human body in each frame of the historical image library, and the vector corresponding to the appearance features of the human body in each frame of the historical image library can be obtained through the appearance features of the human body in each frame of the historical image library; the cosine similarity between the appearance features of the missing person A to be identified and the appearance features of the human body in each frame of the historical image library is calculated through the vector corresponding to the appearance features of the missing person A to be identified and the vector corresponding to the appearance features of the human body in each frame of the historical image library, that is, the first score in the third column.

[0150] S507. Determine a plurality of second scores according to the gait features of the target to be identified and the gait features of the target in each frame of the image, wherein each second score is used to indicate a similarity between the gait features of the target to be identified and the gait features of the target in each frame of the image.

[0151] Each second score represents the similarity between the gait feature of the target to be identified and the gait feature of the target in each frame of image.

[0152] In some embodiments, the similarity between the gait features of the target to be identified and the gait features of the target in each frame of image refers to the cosine similarity between the gait features of the target to be identified and the gait features of the target in each frame of image.

[0153] In some embodiments, the cosine similarity calculation formula between the gait feature of the target to be identified and the gait feature of the target in each frame image is:

[0154]

[0155] Among them, s represents the cosine similarity between the gait features of the target to be identified and the gait features of the target in each frame image, and f i Represents the vector corresponding to the gait feature of the target to be identified, f j Represents the vector corresponding to the gait feature of the target in each frame of the historical image library, ||f i ||||f j || represents the modulus of the vector corresponding to the gait feature of the target to be identified and the vector corresponding to the gait feature of the target in each frame image.

[0156] In a specific embodiment, assuming that the target re-identification device is a mobile phone and the target re-identification apparatus is a mini program, the mini program can display a search page for the target re-identification apparatus. In the missing person search scenario, after the user enters multiple frames of images in which each frame includes the missing person A to be identified in the search page of the mini program, the mini program can extract the gait features of the human body in each frame of the image from the historical image library, and calculate the cosine similarity of the walking movements and postures of the missing person A to be identified and the walking movements and postures of the human body in each frame of the image in the historical image library.

[0157] When there are 100 frames of images in the historical image library, the cosine similarity between the walking motion and posture of the missing person A to be identified and the walking motion and posture of the human body in each frame of the 100 frames of images is calculated.

[0158] like Figure 5c As shown, the first column is the gait features of the missing person A to be identified, and the vector corresponding to the gait features of the missing person A to be identified can be obtained through the gait features of the missing person A to be identified; the second column is the gait features of the human body in each frame of the historical image library, and the vector corresponding to the gait features of the human body in each frame of the historical image library can be obtained through the gait features of the human body in each frame of the historical image library; the cosine similarity between the gait features of the missing person A to be identified and the gait features of the human body in each frame of the historical image library is calculated through the vector corresponding to the gait features of the missing person A to be identified and the vector corresponding to the gait features of the human body in each frame of the historical image library, that is, the second score in the third column.

[0159] S508: Merge the multiple first scores and the multiple second scores to obtain multiple scores.

[0160] In some embodiments, each frame of the image in the historical image library is associated with a first score and a second score, and then the first score and the second score are fused to obtain multiple scores.

[0161] Since the value range of cosine similarity is between -1 and 1, the fusion of cosine similarity cannot simply be achieved by multiplying two cosine similarities.

[0162] In some embodiments, two cosine similarities are fused using the following formula:

[0163] s fusion =fusion(s;λ0,γ0)·fusion(s′;λ1,γ1)

[0164]

[0165] Among them, s fusionrepresents the value of the score obtained by fusing the first score and the second score, fusion represents a conversion function, and the value range of the cosine similarity can be converted to between 0 and 1 through the conversion function, λ and γ represent the parameters of the conversion function, λ0 and γ0 represent the parameters of the conversion function of the cosine similarity between the gait feature and the gait feature of the target in each frame image, λ1 and γ1 represent the parameters of the conversion function of the cosine similarity between the appearance feature and the appearance feature of the target in each frame image.

[0166] In a specific embodiment, assuming that the target re-identification device is a mobile phone and the target re-identification apparatus is a mini program, the mini program can display a search page for the target re-identification apparatus. In the missing person search scenario, after the user enters multiple frames of images each of which includes the missing person A to be identified in the mini program search page, the mini program can merge the first score and the second score corresponding to each frame of the image to obtain multiple scores.

[0167] When there are 100 frames of images in the historical image library, the first score and the second score corresponding to each frame of the 100 frames of images are matched, for example, the first score corresponding to image 1 and the second score corresponding to image 1 are aligned first, and then fused.

[0168] S509. Take the first N scores from the multiple scores in descending order, where N is a positive integer.

[0169] Wherein, N is preset, for example, N may be 50. When N is 50, taking the top N scores from the multiple scores is taking the top 50 scores from the multiple scores.

[0170] In a specific embodiment, assuming that the target re-identification device is a mobile phone and the target re-identification apparatus is a mini program, the mini program can display a search page for the target re-identification apparatus. In the missing person search scenario, after the user enters multiple frames of images in which each frame includes the missing person A to be identified in the mini program search page, the mini program can merge the first score and the second score corresponding to each frame of the image to obtain multiple scores, and sort the score values ​​in descending order, and take the top N scores from them.

[0171] When there are 100 frames of images in the historical image library, the first score and the second score corresponding to each frame of the 100 frames are matched, for example, the first score corresponding to image 1 and the second score corresponding to image 1 are aligned and fused to obtain 100 scores, and the first 50 scores are taken from the 100 scores.

[0172] S510: Obtain images corresponding to the top N scores from the historical image library.

[0173] In a specific embodiment, assuming that the target re-identification device is a mobile phone, and the target re-identification apparatus is a mini program, the mini program can display a search page for the target re-identification apparatus. In the missing person search scenario, after the user enters multiple frames of images in which each frame includes the missing person A to be identified in the mini program search page, the mini program can merge the first score and the second score corresponding to each frame of the image to obtain multiple scores, and sort the score values ​​in descending order, and take the images corresponding to the first N scores.

[0174] When there are 100 frames of images in the historical image library, the first score and the second score corresponding to each frame of the 100 frames are matched, for example, the first score corresponding to image 1 and the second score corresponding to image 1 are aligned and fused to obtain 100 scores, and images corresponding to the first 50 scores are taken from the 100 scores.

[0175] S511 , determining that the images matching the target to be identified are the images corresponding to the top N scores.

[0176] In a specific embodiment, assuming that the target re-identification device is a mobile phone, and the target re-identification apparatus is a mini program, the mini program can display a search page for the target re-identification apparatus. In the missing person search scenario, after the user enters multiple frames of images in which each frame includes the missing person A to be identified in the mini program search page, the mini program can merge the first score and the second score corresponding to each frame of the image to obtain multiple scores, and sort the score values ​​in descending order, and take the images corresponding to the first N scores as images matching the target to be identified, and display them on the mini program search page.

[0177] In the present application, the target re-identification device inputs a single-frame image in a multi-frame image into an appearance feature extraction network to obtain the appearance features of the target to be identified that are independent of the color of the single-frame image, and determines the gait features of the target to be identified through the continuous frame image; multiple first scores are determined according to the appearance features of the target to be identified and the appearance features of the target in each frame image, and multiple second scores are determined according to the gait features of the target to be identified and the gait features of the target in each frame image, and multiple first scores and multiple second scores are fused to obtain multiple scores, and the images corresponding to the first N scores are obtained in descending order as the images matching the target to be identified. By fusing the two scores, and then using the images corresponding to the first N scores obtained by the fusion as the images matching the target to be identified, the effect of improving the recognition accuracy can be achieved; at the same time, using the images corresponding to the first N scores as the images matching the target to be identified, the number of images similar to the target to be identified is large, which can avoid missing images similar to the target to be identified and ensure the humanization of recognition.

[0178] based on Figure 1 According to the description of the illustrated embodiment, when acquiring appearance features, the target re-identification device sequentially inputs each frame of multiple frames into an appearance feature extraction network to obtain multiple appearance features, and then fuses the multiple appearance features.

[0179] Next, combine Figure 6 , the specific implementation method of the target re-identification device to perform the above process is introduced in detail.

[0180] See also Figure 6 , Figure 6 A schematic flow chart of a target re-identification method provided in an embodiment of the present application is shown.

[0181] like Figure 6 As shown, the target re-identification method provided by the present application may include:

[0182] S601, input each frame of the multiple frames of images into the appearance feature extraction network to obtain multiple appearance features, each appearance feature is independent of the color of each corresponding frame of the image.

[0183] Among them, S601 and Figure 1 The implementation method of S102 in the illustrated embodiment is similar and will not be described in detail in this application.

[0184] In some embodiments, the formula for extracting appearance features through the appearance feature extraction network is:

[0185]

[0186]

[0187] Among them, f′ i Represents the multi-frame image K of the target identity i obtained by the conversion function of the appearance feature extraction network i The appearance features of G represent the conversion function of the appearance feature extraction network, K i represents a multi-frame image including the target identity i to be identified, n represents the total number of frames of the multi-frame image, P represents the identity of the target to be identified, P i represents the target to be identified with identity i, represents the jth frame image of the target to be identified with identity i in multiple frames of images, Indicates that, P i The j-th frame image.

[0188] S602: Fusing the multiple appearance features to obtain appearance features of the target to be identified.

[0189] In the present application, the target re-identification device inputs each frame of the multiple frames into the appearance feature extraction network, obtains multiple appearance features respectively, and fuses the multiple appearance features to obtain the appearance features of the target to be identified. By fusing the appearance features corresponding to each frame of the multiple frames to obtain the appearance features of the target to be identified, it can be ensured that the obtained appearance features are more accurate, so as to further ensure the accuracy of recognition.

[0190] Corresponding to the above Figure 1 The illustrated embodiment describes a target re-identification method, and the present application also provides a target re-identification device.

[0191] Next, combine Figure 7 , a target re-identification device provided in an embodiment of the present application is described in detail.

[0192] See also Figure 7 , Figure 7 A schematic block diagram of a target re-identification device provided in an embodiment of the present application is shown.

[0193] like Figure 7 As shown, a target re-identification device provided by an embodiment of the present application includes a first acquisition module 701 , a second acquisition module 702 and a recognition module 703 .

[0194] The first acquisition module 701 is used to acquire multiple frames of images, each of which includes a target to be identified;

[0195] A second acquisition module 702 is used to input a single frame image of the multiple frames into an appearance feature extraction network to obtain the appearance features of the target to be identified, wherein the appearance features of the target to be identified are independent of the color of the single frame image, and the appearance feature extraction network is used to eliminate the interference of the color of the image on the appearance of the target;

[0196] The recognition module 703 is used to determine an image matching the target to be recognized from a historical image library according to the appearance features of the target to be recognized.

[0197] In some embodiments, a feature network generation system is used to:

[0198] Acquire at least one frame of sample image, each frame of the sample image including a sample target;

[0199] Performing data enhancement on each pixel point in each frame of the sample image to obtain multiple frames of enhanced images;

[0200] Determining the appearance features of the sample target according to the multiple frames of enhanced images, where the appearance features of the sample target are independent of the colors of the multiple frames of enhanced images;

[0201] The original feature extraction network is trained according to the appearance features of the sample target to obtain the appearance feature extraction network.

[0202] In some embodiments, the feature network generation system is specifically used to:

[0203] For each pixel point in each frame of the sample image, the color values ​​of each color channel are exchanged and the gray value is converted to obtain multiple frames of enhanced images;

[0204] Alternatively, the color values ​​of each color channel are exchanged for each pixel in each frame of the sample image to obtain multiple frames of enhanced images.

[0205] In some embodiments, the target re-identification device 700 further includes: a third acquisition module ( Figure 7 not indicated).

[0206] The third acquisition module is used to:

[0207] Acquire temporally continuous frame images from the multiple frame images;

[0208] The gait features of the target to be identified are determined according to the continuous frame images, and the gait features of the target to be identified are used to represent the posture and movement of the target to be identified when walking.

[0209] In some embodiments, the identification module 703 is specifically configured to:

[0210] According to the appearance features of the target to be identified and the gait features of the target to be identified, an image matching the target to be identified is determined from the historical image library.

[0211] In some embodiments, the third acquisition module is specifically configured to:

[0212] Separating a foreground and background region from a background region of each frame image in the continuous frame images, wherein the foreground and background region includes the target to be identified, and the background region does not include the target to be identified;

[0213] Using the foreground and background area of ​​each frame image in the continuous frame images as the continuous frame foreground and background image;

[0214] The gait features of the target to be identified are determined according to the background images before the continuous frames.

[0215] In some embodiments, the identification module 703 is specifically configured to:

[0216] In the historical image library, determining the appearance features and gait features of the target in each frame of the image;

[0217] Determining a plurality of first scores according to the appearance features of the target to be identified and the appearance features of the target in each frame of the image, each first score being used to indicate the similarity between the appearance features of the target to be identified and the appearance features of the target in each frame of the image;

[0218] Determining a plurality of second scores according to the gait features of the target to be identified and the gait features of the target in each frame of the image, each second score being used to indicate a similarity between the gait features of the target to be identified and the gait features of the target in each frame of the image;

[0219] Merging the multiple first scores and the multiple second scores to obtain multiple scores;

[0220] In descending order of the scores, take the first N scores from the multiple scores, where N is a positive integer;

[0221] Obtaining images corresponding to the top N scores from the historical image library;

[0222] Determine that the image matching the target to be identified is the image corresponding to the top N scores.

[0223] In some embodiments, the second acquisition module 702 is used to:

[0224] Inputting each frame of the multiple images into the appearance feature extraction network to obtain multiple appearance features, each appearance feature is independent of the color of each corresponding frame of the image;

[0225] The multiple appearance features are fused to obtain the appearance features of the target to be identified.

[0226] It should be understood that the device 700 of the present application can be implemented by an application-specific integrated circuit (ASIC) or a programmable logic device (PLD), and the PLD can be a complex programmable logical device (CPLD), a field-programmable gate array (FPGA), a generic array logic (GAL) or any combination thereof. It can also be implemented by software. Figure 1 The target re-identification method shown in the figure is implemented by software Figure 1 When the target re-identification method is shown, the device 700 and its various modules may also be software modules.

[0227] Figure 8 This is a schematic diagram of the structure of a target re-identification device provided in this application. Figure 8 As shown, the device 800 includes a processor 801, a memory 802, a communication interface 803 and a bus 804. The processor 801, the memory 802 and the communication interface 803 communicate via the bus 804, and can also communicate via other means such as wireless transmission. The memory 802 is used to store instructions, and the processor 801 is used to execute the instructions stored in the memory 802. The memory 802 stores program code 8021, and the processor 801 can call the program code 8021 stored in the memory 802 to execute Figure 1 The target re-identification method shown.

[0228] It should be understood that in the present application, the processor 501 may be a CPU, and the processor 801 may also be other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field programmable gate arrays (FPGAs) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or any conventional processor, etc.

[0229] The memory 802 may include a read-only memory and a random access memory, and provide instructions and data to the processor 801. The memory 802 may also include a non-volatile random access memory. The memory 802 may be a volatile memory or a non-volatile memory, or may include both volatile and non-volatile memories. Among them, the non-volatile memory may be a read-only memory (ROM), a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), an electrically erasable programmable read-only memory (EEPROM), or a flash memory. The volatile memory may be a random access memory (RAM), which is used as an external cache. By way of example and not limitation, many forms of RAM are available, such as static RAM (SRAM), dynamic random access memory (DRAM), synchronous dynamic random access memory (SDRAM), double data rate synchronous dynamic random access memory (DDR SDRAM), enhanced synchronous dynamic random access memory (ESDRAM), synchronous link dynamic random access memory (SLDRAM), and direct rambus RAM (DR RAM).

[0230] In addition to the data bus, the bus 804 may also include a power bus, a control bus, and a status signal bus. Figure 8 Various buses are labeled as bus 804.

[0231] It should be understood that the device 800 according to the present application may correspond to the apparatus 700 in the present application, and may correspond to the apparatus 700 in the present application. Figure 1 In the method shown in the figure, when the device 800 corresponds to Figure 1 When the device in the method shown in the figure is used, the above and other operations and / or functions of each module in the device 800 are respectively to implement Figure 1 The operation steps of the method executed by the device in the embodiment are not repeated here for the sake of brevity.

[0232] The present application also provides a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps in the above-mentioned various method embodiments can be implemented.

[0233] The present application provides a computer program product. When the computer program product runs on a mobile terminal, the mobile device can implement the steps in the above-mentioned various method embodiments when executing the computer program product.

[0234] It should be understood that the size of the serial numbers of the steps in the above embodiments does not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of this application.

[0235] It should be noted that the information interaction, execution process, etc. between the above-mentioned devices / units are based on the same concept as the method embodiment of the present application. Their specific functions and technical effects can be found in the method embodiment part and will not be repeated here.

[0236] The technicians in the relevant field can clearly understand that for the convenience and simplicity of description, only the division of the above-mentioned functional units and modules is used as an example for illustration. In practical applications, the above-mentioned function allocation can be completed by different functional units and modules as needed, that is, the internal structure of the above-mentioned device can be divided into different functional units or modules to complete all or part of the functions described above. The functional units and modules in the embodiment can be integrated in a processing unit, or each unit can exist physically separately, or two or more units can be integrated in one unit. The above-mentioned integrated unit can be implemented in the form of hardware or in the form of software functional units. In addition, the specific names of the functional units and modules are only for the convenience of distinguishing each other, and are not used to limit the scope of protection of this application. The specific working process of the units and modules in the above-mentioned system can refer to the corresponding process in the aforementioned method embodiment, which will not be repeated here.

[0237] In the above embodiments, the description of each embodiment has its own emphasis. For parts that are not described or recorded in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0238] Those of ordinary skill in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.

[0239] In the embodiments provided in the present application, it should be understood that the disclosed devices / network equipment and methods can be implemented in other ways. For example, the device / network equipment embodiments described above are merely schematic. For example, the division of the above modules or units is only a logical function division. There may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0240] The units described above as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the present application.

[0241] The embodiments described above are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the aforementioned embodiments, a person skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features may be replaced by equivalents. Such modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present application, and should all be included in the protection scope of the present application.

Claims

1. A target re-identification method, characterized in that: include: Acquire multiple frames of images, each frame of the image includes the target to be identified; Input a single-frame image in the multiple-frame images into an appearance feature extraction network to obtain the appearance features of the target to be identified, the appearance features of the target to be identified are independent of the color of the single-frame image, the appearance feature extraction network is used to eliminate the interference of the color of the image on the appearance of the target, the appearance feature extraction network is obtained by training an original feature extraction network according to the appearance features of multiple sample targets, the appearance features of the multiple sample targets are obtained according to multiple-frame enhanced images, the multiple-frame enhanced images are obtained by performing a first processing on each pixel point in each frame of the multiple sample images, the first processing includes color value exchange of each color channel, or color value exchange and gray value conversion of each color channel, and the appearance features of the multiple sample targets are independent of the color of the multiple-frame enhanced images; According to the appearance features of the target to be identified, an image matching the target to be identified is determined from a historical image library.

2. The method according to claim 1, characterized in that The method further comprises: Acquire temporally continuous frame images from the multiple frame images; Determining the gait features of the target to be identified according to the continuous frame images, wherein the gait features of the target to be identified are used to represent the posture and movement of the target to be identified when walking; The step of determining an image matching the target to be identified from a historical image library according to the appearance features of the target to be identified includes: According to the appearance features of the target to be identified and the gait features of the target to be identified, an image matching the target to be identified is determined from the historical image library.

3. The method according to claim 2, characterized in that Determining the gait features of the target to be identified according to the continuous frame images includes: Separating a foreground and background region from a background region of each frame image in the continuous frame images, wherein the foreground and background region includes the target to be identified, and the background region does not include the target to be identified; Using the foreground and background area of ​​each frame image in the continuous frame images as the continuous frame foreground and background image; The gait features of the target to be identified are determined according to the background images before the continuous frames.

4. The method according to claim 2 or 3, characterized in that The determining, from the historical image library, an image matching the target to be identified based on the appearance features of the target to be identified and the gait features of the target to be identified, comprises: In the historical image library, determining the appearance features and gait features of the target in each frame of the image; Determining a plurality of first scores according to the appearance features of the target to be identified and the appearance features of the target in each frame of the image, each first score being used to indicate the similarity between the appearance features of the target to be identified and the appearance features of the target in each frame of the image; Determining a plurality of second scores according to the gait features of the target to be identified and the gait features of the target in each frame of the image, each second score being used to indicate a similarity between the gait features of the target to be identified and the gait features of the target in each frame of the image; Merging the multiple first scores and the multiple second scores to obtain multiple scores; In descending order of the scores, take the first N scores from the multiple scores, where N is a positive integer; Obtaining images corresponding to the top N scores from the historical image library; Determine that the image matching the target to be identified is the image corresponding to the top N scores.

5. The method according to any one of claims 1 to 3, characterized in that: The step of inputting a single frame image from the multiple frames of images into an appearance feature extraction network to obtain the appearance features of the target to be identified comprises: Inputting each frame of the multiple images into the appearance feature extraction network to obtain multiple appearance features, each appearance feature being independent of the color of each corresponding frame of the image; The multiple appearance features are fused to obtain the appearance features of the target to be identified.

6. A target re-identification device, characterized in that: include: A first acquisition module is used to acquire multiple frames of images, each of which includes a target to be identified; A second acquisition module is used to input a single-frame image in the multiple-frame images into an appearance feature extraction network to obtain the appearance features of the target to be identified, the appearance features of the target to be identified are independent of the color of the single-frame image, the appearance feature extraction network is used to eliminate the interference of the color of the image on the appearance of the target, the appearance feature extraction network is obtained by training an original feature extraction network according to the appearance features of multiple sample targets, the appearance features of the multiple sample targets are obtained according to multiple-frame enhanced images, the multiple-frame enhanced images are obtained by performing a first processing on each pixel point in each frame of the multiple sample images, the first processing includes exchanging the color values ​​of each color channel, or exchanging the color values ​​of each color channel and converting the grayscale values, and the appearance features of the multiple sample targets are independent of the color of the multiple-frame enhanced images; The recognition module is used to determine an image matching the target to be recognized from a historical image library according to the appearance features of the target to be recognized.

7. A target re-identification device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the method according to any one of claims 1 to 4 is implemented.

8. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the method according to any one of claims 1 to 4 is implemented.

Images