An object re-identification method, device, electronic equipment and object re-identification system

Abstract

Embodiments of the present application provide an object re-identification method and device, an electronic device and an object re-identification system. The method comprises: obtaining a to-be-identified image, the to-be-identified image comprising a to-be-identified object; determining a first quality score of a body region of the to-be-identified object in the to-be-identified image, and determining a second quality score of a face region of the to-be-identified object in the to-be-identified image; determining a first similarity between the body region and a body region of a target object, and determining a second similarity between the face region and a face region of the target object; determining a matching score of the to-be-identified object and the target object according to the first quality score, the second quality score, the first similarity and the second similarity; and determining whether the to-be-identified object and the target object are the same object according to the matching score. In this way, the object re-identification in a more complex scene can be well handled, the generalization performance of the object re-identification is greatly improved, and the accuracy of the object re-identification is further improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of monitoring, in particular to an object re-identification method and device, an electronic device and an object re-identification system. BACKGROUND

[0002] The construction of smart cities and safe cities cannot be separated from monitoring devices scattered in various public places, and with the widespread deployment of monitoring devices, a large number of monitoring devices have generated a large amount of monitoring video. Based on this, how to identify a specific object in the monitoring video is an important technical node of intelligent security. This method based on computer vision technology to determine whether a specific object exists in an image or video sequence is called an object re-identification method.

[0003] The object re-identification method in the related art greatly reduces the accuracy of the specific object re-identification result because the body, clothing and other characteristics of the object are easily repeated. SUMMARY

[0004] The purpose of the embodiments of the present application is to provide an object re-identification method, device, electronic device and object re-identification system to improve the accuracy of object re-identification. The specific technical solutions are as follows:

[0005] In a first aspect, the embodiments of the present application provide an object re-identification method, which comprises:

[0006] obtaining an image to be identified, the image to be identified comprising an object to be identified;

[0007] determining a first quality score of a body region of the object to be identified in the image to be identified, and determining a second quality score of a face region of the object to be identified in the image to be identified;

[0008] determining a first similarity between the body region of the object to be identified and a body region of a target object, and determining a second similarity between the face region of the object to be identified and a face region of the target object;

[0009] determining a matching score of the object to be identified and the target object according to the first quality score, the second quality score, the first similarity and the second similarity;

[0010] determining whether the object to be identified and the target object are the same object according to the matching score.

[0011] Optionally, the step of determining the first quality score of the body region of the object to be identified in the image to be identified, and determining the second quality score of the face region of the object to be identified in the image to be identified comprises:

[0012] detecting the to-be-identified image to obtain a body region of the to-be-identified object in the to-be-identified image, and performing quality evaluation on the body region of the to-be-identified object to obtain a first quality score of the body region of the to-be-identified object;

[0013] detecting the body region of the to-be-identified object;

[0014] if the face region of the to-be-identified object is detected, performing quality evaluation on the face region of the to-be-identified object to obtain a second quality score of the face region of the to-be-identified object;

[0015] if the face region of the to-be-identified object is not detected, determining that the second quality score of the face region of the to-be-identified object is 0.

[0016] Optionally, the step of determining the matching score of the to-be-identified object and the target object according to the first quality score, the second quality score, the first similarity and the second similarity comprises:

[0017] performing normalization processing on the first quality score and the second quality score respectively to obtain a first normalized score and a second normalized score;

[0018] determining the matching score of the to-be-identified object and the target object according to the first normalized score, the second normalized score, the first similarity and the second similarity.

[0019] Optionally, the step of determining the matching score of the to-be-identified object and the target object according to the first quality score, the second quality score, the first similarity and the second similarity comprises:

[0020] determining a first weight of the body region and a second weight of the face region of the to-be-identified object according to the first quality score and the second quality score, wherein the first quality score is positively correlated with the first weight, and the second quality score is positively correlated with the second weight;

[0021] performing weighted summation on the first similarity and the second similarity by using the first weight and the second weight to obtain the matching score of the to-be-identified object and the target object.

[0022] Optionally, the step of determining the first weight of the body region and the second weight of the face region of the to-be-identified object according to the first quality score and the second quality score comprises:

[0023] if the first quality score is higher than or equal to the preset body score threshold and the second quality score is higher than or equal to the preset face score threshold, determining a ratio of the first quality score and the second quality score to a score sum value, obtaining a first weight of a body region and a second weight of a face region of the to-be-identified object, the score sum value being a sum of the first quality score and the second quality score;

[0024] if the first quality score is lower than the preset body score threshold and the second quality score is higher than or equal to the preset face score threshold, updating the first quality score to 0, determining a ratio of the updated first quality score and the second quality score to the score sum value, and obtaining the first weight and the second weight;

[0025] if the first quality score is higher than or equal to the preset body score threshold and the second quality score is lower than the preset face score threshold, updating the second quality score to 0, determining a ratio of the first quality score and the updated second quality score to the score sum value, and obtaining the first weight and the second weight.

[0026] Optionally, the method further comprises:

[0027] if the first quality score is lower than the preset body score threshold and the second quality score is lower than the preset face score threshold, rejecting to determine whether the to-be-identified object and the target object are the same object.

[0028] Optionally, the step of determining the first quality score of the body region of the to-be-identified object in the to-be-identified image and determining the second quality score of the face region of the to-be-identified object in the to-be-identified image comprises:

[0029] performing weighted average processing on scores of a plurality of preset quality factors of the body region of the to-be-identified object in the to-be-identified image to obtain the first quality score, and performing weighted average processing on scores of the plurality of preset quality factors of the face region of the to-be-identified object in the to-be-identified image to obtain the second quality score; or

[0030] inputting the body region and the face region of the to-be-identified object in the to-be-identified image into a preset quality evaluation model respectively to obtain the first quality score and the second quality score.

[0031] In a second aspect, an object re-identification device is provided, and the device comprises:

[0032] an acquisition module, configured to acquire a to-be-identified image, the to-be-identified image comprising a to-be-identified object;

[0033] The first determining module is configured to determine a first quality score of a body region of the to-be-identified object in the to-be-identified image, and determine a second quality score of a face region of the to-be-identified object in the to-be-identified image.

[0034] The second determining module is configured to determine a first similarity between the body region of the to-be-identified object and a body region of a target object, and determine a second similarity between the face region of the to-be-identified object and a face region of the target object.

[0035] The third determining module is configured to determine a matching score of the to-be-identified object and the target object according to the first quality score, the second quality score, the first similarity and the second similarity.

[0036] The fourth determining module is configured to determine whether the to-be-identified object and the target object are the same object according to the matching score.

[0037] Optionally, the first determining module comprises:

[0038] The first determining submodule is configured to detect the to-be-identified image to obtain the body region of the to-be-identified object in the to-be-identified image, and perform quality evaluation on the body region of the to-be-identified object to obtain the first quality score of the body region of the to-be-identified object.

[0039] The detecting submodule is configured to detect the body region of the to-be-identified object.

[0040] The second determining submodule is configured to, if the face region of the to-be-identified object is detected, perform quality evaluation on the face region of the to-be-identified object to obtain the second quality score of the face region of the to-be-identified object.

[0041] The third determining submodule is configured to, if the face region of the to-be-identified object is not detected, determine that the second quality score of the face region of the to-be-identified object is 0.

[0042] Optionally, the third determining module comprises:

[0043] The first normalization submodule is configured to perform normalization processing on the first quality score and the second quality score respectively to obtain a first normalized score and a second normalized score.

[0044] The fourth determining submodule is configured to determine a matching score of the to-be-identified object and the target object according to the first normalized score, the second normalized score, the first similarity and the second similarity.

[0045] Optionally, the third determining module comprises:

[0046] a fifth determining sub-module, configured to determine a first weight of a body region and a second weight of a face region of the to-be-identified object according to the first quality score and the second quality score, wherein the first quality score is positively correlated with the first weight, and the second quality score is positively correlated with the second weight;

[0047] a sixth determining sub-module, configured to perform weighted summation on the first similarity and the second similarity by using the first weight and the second weight to obtain a matching score of the to-be-identified object and the target object.

[0048] Optionally, the fifth determining sub-module is specifically configured to:

[0049] if the first quality score is higher than or equal to a preset body score threshold value and the second quality score is higher than or equal to a preset face score threshold value, then a ratio of the first quality score and the second quality score to a score sum value is determined to obtain the first weight of the body region and the second weight of the face region of the to-be-identified object, wherein the score sum value is a sum of the first quality score and the second quality score;

[0050] if the first quality score is lower than the preset body score threshold value and the second quality score is higher than or equal to the preset face score threshold value, then the first quality score is updated to 0, and a ratio of the updated first quality score and the second quality score to the score sum value is determined to obtain the first weight and the second weight;

[0051] if the first quality score is higher than or equal to the preset body score threshold value and the second quality score is lower than the preset face score threshold value, then the second quality score is updated to 0, and a ratio of the first quality score and the updated second quality score to the score sum value is determined to obtain the first weight and the second weight.

[0052] Optionally, the fifth determining sub-module is further configured to: if the first quality score is lower than the preset body score threshold value and the second quality score is lower than the preset face score threshold value, then it is determined that the to-be-identified object and the target object are not the same object.

[0053] Optionally, the first determining module is specifically configured to:

[0054] perform weighted average processing on scores of a plurality of preset quality factors of the body region of the to-be-identified object in the to-be-identified image to obtain the first quality score, and perform weighted average processing on scores of the plurality of preset quality factors of the face region of the to-be-identified object in the to-be-identified image to obtain the second quality score; or

[0055] The body region and the face region of the to-be-identified object in the to-be-identified image are respectively input into a preset quality evaluation model to obtain a first quality score and a second quality score.

[0056] In a third aspect, an electronic device is provided, including a processor and a machine readable storage medium, the machine readable storage medium stores machine executable instructions which can be executed by the processor, and the processor is prompted by the machine executable instructions to implement the steps of the object re-identification method.

[0057] In a fourth aspect, an object re-identification system is provided, including a camera and an electronic device.

[0058] The camera is configured to capture a to-be-identified image, and the to-be-identified image includes a to-be-identified object.

[0059] The electronic device is configured to determine a first quality score of a body region of the to-be-identified object in the to-be-identified image, and determine a second quality score of a face region of the to-be-identified object in the to-be-identified image; determine a first similarity between the body region of the to-be-identified object and a body region of a target object, and determine a second similarity between the face region of the to-be-identified object and a face region of the target object; determine a matching score of the to-be-identified object and the target object according to the first quality score, the second quality score, the first similarity and the second similarity; and determine whether the to-be-identified object and the target object are the same object according to the matching score.

[0060] In a fifth aspect, a computer readable storage medium is provided, and the computer readable storage medium stores a computer program product, and the computer program product is executed by a processor to implement the steps of the object re-identification method.

[0061] In a sixth aspect, a computer program product is provided, and the computer program product is executed by a processor to implement the steps of the object re-identification method.

[0062] The embodiments of the present application have the following beneficial effects:

[0063] In the technical scheme provided in the embodiments of the present application, the first quality score of the body region of the to-be-identified object, the first similarity between the body region of the to-be-identified object and the body region of the target object, the second quality score of the face region of the to-be-identified object, and the second similarity between the face region of the to-be-identified object and the face region of the target object are used to determine whether the to-be-identified object and the target object are the same object, thereby completing object re-identification. In the case of comprehensively considering the information of the body region and the information of the face region, the object re-identification false positives and false negatives caused by the change of the information of the body region or the poor image quality can be effectively avoided, and the accuracy of object re-identification is improved.

[0064] In addition, face recognition is not sensitive to complex scenes such as object dressing scenes and scenes in which the hairstyle or external decoration changes, and the robustness of face recognition to complex scenes such as large poses, low image quality, strong / weak attention, and the like is good, and the generalization performance of face recognition in infrared scenes is good. Therefore, the technical scheme provided in the embodiments of the present application uses the second quality score and the second similarity obtained by face recognition in combination with the first quality score and the first similarity obtained by body recognition, which can well cope with object re-identification in more complex scenes, greatly improves the generalization performance of object re-identification, and further improves the accuracy of object re-identification.

[0065] Of course, implementing any product or method of the present application does not necessarily require all the advantages described above. BRIEF DESCRIPTION OF DRAWINGS

[0066] In order to more clearly illustrate the technical schemes in the embodiments of the present application or the prior art, the drawings needed in the embodiments or the prior art description will be briefly introduced below. Obviously, the drawings in the following description are only some embodiments of the present application, and other embodiments can also be obtained by those skilled in the art based on these drawings.

[0067] Figure 1 The first flowchart of the object re-identification method provided in the embodiments of the present application;

[0068] Figure 2 The second flowchart of the object re-identification method provided in the embodiments of the present application;

[0069] Figure 3 The third flowchart of the object re-identification method provided in the embodiments of the present application;

[0070] Figure 4 The fourth flowchart of the object re-identification method provided in the embodiments of the present application;

[0071] Figure 5 The fifth flowchart of the object re-identification method provided in the embodiments of the present application;

[0072] Figure 6 A structural schematic diagram of an object re-identification device provided by an embodiment of the present application;

[0073] Figure 7 A structural schematic diagram of an electronic device provided by an embodiment of the present application;

[0074] Figure 8 A structural schematic diagram of an object re-identification system provided by an embodiment of the present application. DETAILED DESCRIPTION

[0075] The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art based on the present application belong to the scope of protection of the present application.

[0076] Object re-identification: also known as object re-recognition, is a technology of judging whether a specific object exists in an image or a video sequence by using computer vision technology.

[0077] The object re-identification method in the related art focuses on the features of the body, clothing, etc. of the object. However, in complex scenes such as object dressing scenes and scenes where the hairstyle or exterior decoration changes, the features of the body, clothing, etc. of the object are easy to repeat, which greatly reduces the accuracy of the object re-identification result.

[0078] In order to improve the accuracy of the re-identification result, an object re-identification method is provided in the embodiments of the present application. The object re-identification method can be applied to a monitoring device or an electronic device connected with the monitoring device. For the convenience of understanding, the electronic device is taken as the execution subject for description below, which does not serve as a limitation.

[0079] In the object re-identification method, the first quality score of the body region of the to-be-identified object, the first similarity of the body region of the to-be-identified object and the body region of the target object, the second quality score of the face region of the to-be-identified object, and the second similarity of the face region of the to-be-identified object and the face region of the target object are used to determine whether the to-be-identified object and the target object are the same object, and to complete the re-identification of the target object. In the case of comprehensively considering the information of the body region and the information of the face region, the false positives and false negatives of the object re-identification caused by the change of the information of the body region or the poor image quality can be effectively avoided, and the accuracy of the object re-identification is improved.

[0080] Furthermore, facial recognition is insensitive to complex scenarios such as changes in clothing, hairstyle, or accessories. It also exhibits good robustness in complex scenarios involving large poses, low image quality, and strong / weak illumination, and demonstrates good generalization performance in infrared scenes. Therefore, the object re-identification method provided in this application, by combining the second quality score and second similarity obtained from facial recognition with the first quality score and first similarity obtained from body recognition, can effectively handle object re-identification in more complex scenarios, significantly improving the generalization performance and accuracy of object re-identification.

[0081] The object re-identification method provided in this application will be described in detail below through specific embodiments.

[0082] like Figure 1 As shown, Figure 1 This is a schematic diagram of a first-order object re-identification method provided in an embodiment of this application. The object re-identification method includes the following steps.

[0083] Step S101: Obtain the image to be recognized, which includes the object to be recognized.

[0084] The objects can be pedestrians or animals, etc.

[0085] In this embodiment, the image to be identified can be extracted from video stream data. For example, an electronic device acquires video stream data, identifies each video frame in the video stream data, and obtains a video frame containing the object to be identified, i.e., the image to be identified. The video stream data can be data collected in real time by a monitoring device, or data pre-stored in the electronic device; there is no limitation on this.

[0086] In this embodiment of the application, the image to be identified may also be an image pre-stored in an electronic device. That is, the electronic device can directly obtain the pre-stored image, which is used as the image to be identified, and the object included in the image is the object to be identified.

[0087] Step S102: Determine the first quality score of the body region of the object to be identified in the image to be identified, and determine the second quality score of the face region of the object to be identified in the image to be identified.

[0088] In this embodiment, the electronic device detects the image to be identified, thereby obtaining the body region and face region of the object to be identified. After determining the body region and face region of the object to be identified, the electronic device performs a quality assessment on the detected body region to obtain a first quality score; and performs a quality assessment on the detected face region to obtain a second quality score.

[0089] In an optional embodiment, the electronic device performs weighted average processing on scores of a plurality of preset quality factors of the body region of the to-be-identified object to obtain a first quality score, and performs weighted average processing on scores of a plurality of preset quality factors of the face region of the to-be-identified object to obtain a second quality score.

[0090] In the embodiments of the present application, the electronic device determines scores of each preset quality factor in the body region of the to-be-identified object, and performs weighted average processing on the obtained scores to obtain a first quality score of the body region of the to-be-identified object. In addition, the electronic device determines scores of each preset quality factor in the face region of the to-be-identified object, and performs weighted average processing on the obtained scores to obtain a second quality score of the face region of the to-be-identified object. The preset quality factors can include, but are not limited to, target posture, clarity, illumination degree, and occlusion degree, etc.

[0091] In another optional embodiment, a quality evaluation model is pre-configured in the electronic device, that is, a preset quality evaluation model. The electronic device inputs the body region and the face region of the to-be-identified object in the to-be-identified image into the preset quality evaluation model to obtain the first quality score and the second quality score. Specifically, the body region of the to-be-identified object can be input into the preset quality evaluation model, and the preset quality evaluation model can process the body region of the to-be-identified object to output the first quality score of the body region of the to-be-identified object. The face region of the to-be-identified object can be input into the preset quality evaluation model, and the preset quality evaluation model can process the face region of the to-be-identified object to output the second quality score of the face region of the to-be-identified object.

[0092] The preset quality evaluation model can be a neural network model or a deep neural network model, and is not limited in this regard.

[0093] In the embodiments of the present application, the preset quality evaluation model is a model obtained by training a preset neural network model according to preset training data. The preset training data includes sample images and corresponding sample quality scores. In order to improve the accuracy of the preset quality evaluation model obtained by training, the sample images can include images of body regions and images of face regions.

[0094] The training process of the preset quality evaluation model can include:

[0095] The electronic device obtains preset training data, inputs a sample image into a preset neural network model to obtain a predicted quality score of the sample image, determines a loss value of quality evaluation according to the predicted quality score and a sample quality score, and determines whether the preset neural network model converges according to the loss value; if the preset neural network model converges, the training is ended, and the current preset neural network model is taken as a preset quality evaluation model; if the preset neural network model does not converge, the parameters of the preset neural network model are adjusted, and the step of inputting the sample image into the preset neural network model to obtain the predicted quality score of the sample image is executed again, and the iteration training is continued until the preset neural network model converges.

[0096] In step S103, the first similarity between the body region of the to-be-identified object and the body region of the target object is determined, and the second similarity between the face region of the to-be-identified object and the face region of the target object is determined.

[0097] In the embodiment of the present application, the body region and the face region of the target object are determined in advance in the electronic device. After the body region and the face region of the to-be-identified object are determined, the electronic device can compare the body region of the to-be-identified object with the body region of the target object to obtain the first similarity between the body region of the to-be-identified object and the body region of the target object, and compare the face region of the to-be-identified object with the face region of the target object to obtain the second similarity between the face region of the to-be-identified object and the face region of the target object.

[0098] In the embodiment of the present application, the electronic device can use a feature extraction algorithm such as a deep neural network, a SIFT (Scale Invariant Feature Transform) algorithm or a HOG (Histogram Of Gradient) algorithm to extract the first body feature of the body region of the to-be-identified object, the first face feature of the face region of the to-be-identified object, the second body feature of the body region of the target object, and the second face feature of the face region of the target object.

[0099] The electronic device determines the first similarity between the body region of the to-be-identified object and the body region of the target object based on the first body feature and the second body feature, and determines the second similarity between the face region of the to-be-identified object and the face region of the target object based on the first face feature and the second face feature.

[0100] In the embodiment of the present application, the deep neural network for extracting features in the electronic device includes but is not limited to ResNet (Residual Neural Network), AlexNet, VGGNet and Transformer.

[0101] Step S104, according to the first quality score, the second quality score, the first similarity and the second similarity, a matching score of the to-be-identified object and the target object is determined.

[0102] In the embodiment of the present application, after the first quality score of the body region of the to-be-identified object, the second quality score of the face region of the to-be-identified object, the first similarity score of the body region of the to-be-identified object and the body region of the target object, and the second similarity score of the face region of the to-be-identified object and the face region of the target object, the first quality score and the second quality score can be used to fuse the first similarity and the second similarity, and the matching score of the to-be-identified object and the target object is obtained.

[0103] Step S105, according to the matching score, whether the to-be-identified object and the target object are the same object is determined.

[0104] In the embodiment of the present application, the electronic device can set a matching degree threshold, that is, a preset matching degree threshold. If the matching score of the to-be-identified object and the target object is greater than the preset matching degree threshold, the electronic device can determine that the target object matches the to-be-identified object, that is, the to-be-identified object and the target object are the same object. If the matching score of the to-be-identified object and the target object is less than or equal to the preset matching degree threshold, the electronic device can determine that the target object does not match the to-be-identified object, that is, the to-be-identified object and the target object are not the same object.

[0105] In one embodiment, the number of target objects is multiple. If the matching scores of the to-be-identified object and the multiple target objects are all greater than the preset matching degree threshold, the electronic device can take the target object with the largest matching score with the to-be-identified object as the target object matching the to-be-identified object.

[0106] In one embodiment, the to-be-identified image includes multiple to-be-identified objects. If the matching scores of the multiple to-be-identified objects and the target object are all greater than the preset matching degree threshold, the electronic device can sort the multiple to-be-identified objects according to the matching scores from high to low, and output the sorted results, so that the user can determine the to-be-identified object matching the target object from the sorted results, that is, determine the to-be-identified object which is the same object as the target object from the sorted results.

[0107] In the embodiment of the present application, the electronic device can pre-store the correspondence between the target object and the target operation. After determining that the to-be-identified object and the target object are the same object, the electronic device can obtain the target operation corresponding to the target object, and then execute the corresponding target operation. The target operation can be opening a door, paying, or unlocking a mobile phone, etc.

[0108] Taking a door access system of a community as an example, a pedestrian is taken as the to-be-identified object. The door access system includes an electronic device in the background and a camera at the entrance of the community. The camera at the entrance of the community collects a real-time to-be-identified image, and the electronic device performs object re-identification on the to-be-identified image according to the technical solution provided in the present application example, to determine whether the pedestrian in the to-be-identified image exists in the pre-stored objects, that is, to determine whether the pedestrian in the to-be-identified image is a resident in the community. If the result of object re-identification is that the pre-stored target object and the pedestrian in the to-be-identified image are the same object, it can be determined that the pedestrian is a resident in the community, and the door access is opened to enable the pedestrian to enter the community; if the result of object re-identification is that the pre-stored object does not exist in the pedestrian in the to-be-identified image, it can be determined that the pedestrian is not a resident in the community, and the door access is kept closed to prohibit non-community personnel from entering the community, thereby improving the security of the community.

[0109] In the technical solution provided in the present application example, the first quality score of the body region of the to-be-identified object, the first similarity of the body region of the target object in the body region of the to-be-identified object, the second quality score of the face region of the to-be-identified object, and the second similarity of the face region of the target object in the face region of the to-be-identified object are used to determine whether the to-be-identified object and the target object are the same object, and to complete the re-identification of the target object. In the case of comprehensively considering the information of the body region and the information of the face region, the false positives and false negatives of object re-identification caused by changes in the information of the body region or poor image quality can be effectively avoided, and the accuracy of object re-identification is improved.

[0110] In addition, face recognition is not sensitive to complex scenes such as object dressing and changes in hairstyle or accessories, and has good robustness to complex scenes such as large poses, low image quality, strong / weak attention, and good generalization performance in infrared scenes. Therefore, the technical solution provided in the present application example uses the second quality score and the second similarity obtained by face recognition, in combination with the first quality score and the first similarity obtained by body recognition, to well cope with object re-identification in more complex scenes, greatly improves the generalization performance of object re-identification, and further improves the accuracy of object re-identification.

[0111] In an embodiment of the present application, the present application example further provides an object re-identification method, as shown in Figure 2 In the method, step S102 can be refined into steps S1021 to S1024.

[0112] Step S1021 detects the to-be-identified image to obtain the body region of the to-be-identified object in the to-be-identified image, and performs quality evaluation on the body region of the to-be-identified object to obtain the first quality score of the body region of the to-be-identified object.

[0113] The electronic device detects the to-be-recognized image to obtain a body region of the to-be-recognized object. Quality assessment is performed on the body region of the to-be-recognized object to obtain a first quality score of the body region of the to-be-recognized object. The quality assessment method can be any quality assessment method in the related art, and is not limited in this regard.

[0114] In the embodiments of the present application, the quality assessment of the body region of the to-be-recognized object by the electronic device can consider the first image dimension of the to-be-recognized image. The first image dimension includes but is not limited to body posture, degree of body occlusion, degree of clarity of the picture, and degree of illumination of the picture.

[0115] Step S1022, detecting the body region of the to-be-recognized object.

[0116] After determining the body region of the to-be-recognized object, the electronic device can detect the body region of the to-be-recognized object to determine whether the face region of the to-be-recognized object exists in the body region of the to-be-recognized object. If the face region of the to-be-recognized object is detected, i.e., the face region of the to-be-recognized object exists and the face region is not completely occluded, step S1023 is performed; if the face region of the to-be-recognized object is not detected, i.e., the face region of the to-be-recognized object does not exist and the face region is completely occluded, step S1024 is performed.

[0117] Step S1023, quality assessment is performed on the face region of the to-be-recognized object to obtain a second quality score of the face region of the to-be-recognized object.

[0118] After detecting the body region of the to-be-recognized object, the electronic device performs quality assessment on the face region of the to-be-recognized object if the face region of the to-be-recognized object is obtained, and a second quality score of the body region of the to-be-recognized object can be obtained.

[0119] In the embodiments of the present application, the quality assessment of the face region of the to-be-recognized object by the electronic device can consider the second image dimension of the to-be-recognized image. The second image dimension includes but is not limited to face posture, degree of face occlusion, degree of clarity of the picture, and degree of illumination of the picture.

[0120] Step S1024, determining that the second quality score of the face region of the to-be-recognized object is 0.

[0121] The face region of the to-be-identified object can not be detected due to serious occlusion or too dark light, and the like. In this case, the electronic device detects the body region of the to-be-identified object. If the face region of the to-be-identified object is not detected in the body region of the to-be-identified object, it is determined that the face region of the to-be-identified object does not exist. In this case, the quality assessment of the face region of the to-be-identified object is rejected, and the second quality score of the face region of the to-be-identified object is directly determined as 0, thereby improving the object re-identification efficiency.

[0122] In the technical scheme provided in the embodiments of the present application, the electronic device detects the to-be-identified image, first acquires the body region of the to-be-identified object image, and then detects whether the face region of the to-be-identified object exists in the body region of the to-be-identified object. In this way, the detection efficiency of the face region of the to-be-identified object can be improved.

[0123] In addition, in the technical scheme provided in the embodiments of the present application, in the case that there are multiple to-be-identified objects in one to-be-identified image, the face region of a to-be-identified object is acquired from the body region of the to-be-identified object, and the first quality score and the second quality score of the to-be-identified object are determined. This can effectively avoid the problem that the body region and the face region of different to-be-identified objects are determined as the body region and the face region of the same to-be-identified object, thereby improving the accuracy of object re-identification.

[0124] In one embodiment of the present application, the embodiments of the present application further provide an object re-identification method, as shown in Figure 3 In the method, step S104 can be refined into step S1041 and step S1042.

[0125] Step S1041, respectively normalizing the first quality score and the second quality score to obtain a first normalized score and a second normalized score.

[0126] In the embodiments of the present application, the electronic device can normalize the first quality score to obtain a first normalized score, and normalize the second quality score to obtain a second normalized score. The first normalized score and the second normalized score are under the same standard, for example, the value range of the first normalized score and the second normalized score is between 0 and 1.

[0127] In the embodiments of the present application, the normalization processing mode can adopt a deviation standardization or a standard deviation standardization method, which is not limited.

[0128] Step S1042, determining a matching score of the to-be-identified object and the target object according to the first normalized score, the second normalized score, the first similarity, and the second similarity.

[0129] In the embodiment of the present application, the electronic device can determine the matching score of the to-be-identified object and the target object according to the first normalized score and the second normalized score, in combination with the first similarity between the body region of the to-be-identified object and the body region of the target object, and the second similarity between the face region of the to-be-identified object and the face region of the target object.

[0130] In the technical solution provided by the embodiment of the present application, the electronic device performs normalization processing on the first quality score and the second quality score, and then determines the matching score of the to-be-identified object and the target object by using the first normalized score and the second normalized score obtained after the processing. In this way, the first normalized score and the second normalized score obtained after the normalization processing of the first quality score and the second quality score are under the same standard, so that the first quality score and the second quality score obtained by respectively performing quality evaluation on the body region and the face region of the to-be-identified object are comparable, which is beneficial to the processing of data. In addition, the data after the normalization processing is in a fixed and small value range, which can effectively reduce the calculation amount of the electronic device and improve the efficiency of object re-identification.

[0131] In one embodiment of the present application, the embodiment of the present application further provides an object re-identification method, as shown in Figure 4 As shown in the method, step S104 can be refined into step S1043 and step S1044.

[0132] In step S1043, the first weight of the body region and the second weight of the face region of the to-be-identified object are determined according to the first quality score and the second quality score, wherein the first quality score is positively correlated with the first weight, and the second quality score is positively correlated with the second weight.

[0133] In the embodiment of the present application, the electronic device can determine the first weight positively correlated with the first quality score of the body region of the to-be-identified object and the second weight positively correlated with the second quality score of the face region of the to-be-identified object according to the first quality score of the body region and the second quality score of the face region of the to-be-identified object.

[0134] In one optional embodiment, the electronic device can determine the first weight of the body region and the second weight of the face region of the to-be-identified object by using the following formula.

[0135]

[0136]

[0137] wherein W1 represents a first weight of the body region of the to-be-identified object, W2 represents a second weight of the face region of the to-be-identified object, S1 represents a first quality score of the body region of the to-be-identified object, and S2 represents a second quality score of the face region of the to-be-identified object.

[0138] In an embodiment of the present application, in order to reduce the calculation amount of the electronic device and improve the efficiency of object re-identification, the electronic device performs normalization processing on the first quality score and the second quality score. In this case, in the above formula (1) and formula (2), S1 can represent a first normalized score, and S2 can represent a second normalized score.

[0139] In another optional embodiment, the electronic device can pre-store a corresponding relationship between a quality score range and a weight. After determining the first quality score and the second quality score, the electronic device can find a first quality score range to which the first quality score belongs and a second quality score range to which the second quality score belongs; and determine, according to the pre-stored corresponding relationship between the quality score range and the weight, a weight corresponding to the first quality score range as the first weight and a weight corresponding to the second quality score range as the second weight.

[0140] In an embodiment of the present application, the electronic device can also determine the first weight and the second weight in other manners, which are not limited herein.

[0141] In step S1044, the first weight and the second weight are used to perform weighted summation on the first similarity and the second similarity to obtain a matching score of the to-be-identified object and the target object.

[0142] In an embodiment of the present application, the first weight of the body region of the to-be-identified object can represent the proportion of the body region of the to-be-identified object in the matching score of the to-be-identified object and the target object. That is, the first weight can represent the influence of the first similarity on the matching score. The second weight of the face region of the to-be-identified object can represent the proportion of the face region of the to-be-identified object in the matching score of the to-be-identified object and the target object. That is, the second weight can represent the influence of the second similarity on the matching score.

[0143] In an embodiment of the present application, the electronic device can obtain the matching score of the to-be-identified object and the target object by using the first weight and the second weight to perform weighted summation on the first similarity and the second similarity.

[0144] In an optional embodiment, the electronic device can use the following formula to determine the matching score of the to-be-identified object and the target object.

[0145] sim=W1f1+W2f2 (3)

[0146] Where sim represents the matching score between the object to be identified and the target object, W1 represents the first weight, W2 represents the second weight, f1 represents the first similarity, and f2 represents the second similarity.

[0147] In the technical solution provided in this application embodiment, the higher the first quality score, the greater the positive correlation with its first weight, and the greater the impact of the corresponding first similarity on the matching score. Similarly, the higher the second quality score, the greater the positive correlation with its second weight, and the greater the impact of the corresponding second similarity on the matching score. By utilizing the influence of the first and second weights on the matching scores of the object to be identified and the target object, the electronic device can better combine the body and facial regions of the object to be identified to perform object re-identification, thereby improving the accuracy of object re-identification.

[0148] based on Figure 4 The object re-identification method shown in this application embodiment also provides an object re-identification method, such as... Figure 5 As shown. In this method, step S1043 can be further refined into steps S1043a to S1043c.

[0149] Step S1043a: If the first quality score is higher than or equal to the preset body score threshold and the second quality score is higher than or equal to the preset face score threshold, then determine the ratio of the first quality score and the second quality score to the score sum value, respectively, to obtain the first weight of the body region and the second weight of the face region of the object to be identified, and the score sum value is the sum of the first quality score and the second quality score.

[0150] In this embodiment, the electronic device can preset a body scoring threshold and a face scoring threshold, i.e., a preset body scoring threshold and a preset face scoring threshold. If it is determined that the first quality score is higher than or equal to the preset body scoring threshold and the second quality score is higher than or equal to the preset face scoring threshold, the electronic device can obtain the first weight of the body region of the object to be identified based on the first quality score of the body region of the object to be identified, and obtain the second weight of the face region of the object to be identified based on the second quality score of the face region of the object to be identified. Wherein, the first weight is the ratio of the first quality score to the sum of the scores, the second weight is the ratio of the second quality score to the sum of the scores, and the sum of the scores is the sum of the first quality score and the second quality score. For example, the electronic device can determine the first weight and the second weight based on the first quality score and the second quality score, using the above method (1) and formula (2).

[0151] In this embodiment, the preset body score threshold and the preset facial score threshold can be set according to actual needs. For example, the preset body score threshold can be 80, 90, or 95, etc.

[0152] Step S1043b, if the first quality score is lower than the preset body score threshold and the second quality score is higher than or equal to the preset face score threshold, the first quality score is updated to 0, and the ratio of the updated first quality score and the second quality score to the score sum is determined to obtain the first weight and the second weight.

[0153] In the embodiment of the application, if the first quality score of the body region of the to-be-identified object is lower than the preset body score threshold and the second quality score of the face region of the to-be-identified object is higher than or equal to the preset face score threshold in the electronic device, the first quality score is updated to 0, the second quality score remains unchanged, and the score sum of the updated first quality score and the second quality score is equal to the second quality score. According to the ratio of the updated first quality score to the score sum, the first weight can be determined, and according to the ratio of the second quality score to the score sum, the second weight can be determined.

[0154] For example, the electronic device can determine the first weight and the second weight by using the above method (1) and formula (2) based on the updated first quality score and the second quality score. In the embodiment of the application, if the first quality score of the body region of the to-be-identified object is lower than the preset body score threshold and the second quality score of the face region of the to-be-identified object is higher than or equal to the preset face score threshold, the electronic device updates the first quality score S1 to 0, and the second quality score S2 remains unchanged. The score sum (S1+S2) of the updated first quality score and the second quality score is equal to the second quality score S2. At this time, the electronic device can determine that the first weight W1 is 0 and the second weight W2 is 1.

[0155] Step S1043c, if the first quality score is higher than or equal to the preset body score threshold and the second quality score is lower than the preset face score threshold, the second quality score is updated to 0, and the ratio of the first quality score and the updated second quality score to the score sum is determined to obtain the first weight and the second weight.

[0156] In the embodiment of the application, if the first quality score of the body region of the to-be-identified object is higher than or equal to the preset body score threshold and the second quality score of the face region of the to-be-identified object is lower than the preset face score threshold in the electronic device, the second quality score is updated to 0, the first quality score remains unchanged, and the score sum of the updated first quality score and the second quality score is equal to the first quality score. According to the ratio of the first quality score to the score sum, the first weight can be determined, and according to the ratio of the updated second quality score to the score sum, the second weight can be determined.

[0157] For example, the electronic device can determine the first weight and the second weight based on the first quality score and the updated second quality score, in the above manner (1) and formula (2). In an embodiment of the present application, if the first quality score of the body region of the to-be-identified object is higher than or equal to the preset body score threshold, and the second quality score of the face region of the to-be-identified object is lower than the preset face score threshold, the electronic device keeps the first quality score S1 unchanged, and updates the second quality score S2 to 0. The sum (S1+S2) of the first quality score and the updated second quality score is equal to the first quality score S1. At this time, it can be determined that the first weight W1 is 1, and the second weight W2 is 0.

[0158] In the technical solution provided in the embodiments of the present application, the electronic device determines the first weight and the second weight according to whether the first quality score and the second quality score are lower than the threshold, which can avoid the body region of the to-be-identified object and the face region of the to-be-identified object with too low quality from having a great influence on the matching score of the to-be-identified object and the target object, and causing the false negative or false positive of object re-identification.

[0159] In an embodiment of the present application, if the first quality score of the body region of the to-be-identified object is lower than the preset body score threshold, and the second quality score of the face region of the to-be-identified object is lower than the preset face score threshold, it indicates that the quality of the to-be-identified image is too low, and the electronic device can refuse to identify whether the to-be-identified image and the target image are the same object.

[0160] For example, if the body of the to-be-identified object is seriously occluded, and the face of the to-be-identified object is seriously occluded, resulting in that the first quality score of the body region of the to-be-identified object is lower than the preset body score threshold, and the second quality score of the face region of the to-be-identified object is lower than the preset face score threshold. The quality of the to-be-identified image is too low, and the electronic device can refuse to determine whether the to-be-identified object and the target object are the same object.

[0161] Similarly, if the clarity of the body region and the face region of the to-be-identified object is too low, resulting in that the first quality score is lower than the preset body threshold, and the second quality score is lower than the preset face threshold. The electronic device can refuse to determine whether the to-be-identified object and the target object are the same object.

[0162] The quality evaluation of the body region and the face region of the to-be-identified object is only an example, and is not limited thereto.

[0163] The technical scheme provided in the embodiments of the present application considers the body region and the face region of the to-be-identified object when performing object re-identification. If the first quality score of the body region of the to-be-identified object and the second quality score of the face region of the to-be-identified object are too low, the electronic device refuses to identify whether the to-be-identified object and the target object are the same object. In this way, the problem of object re-identification false positives caused by too low image quality of the to-be-identified object is solved, the computing resources of the electronic device are saved, and the re-identification efficiency is improved.

[0164] Corresponding to the object re-identification method described above, the embodiments of the present application provide an object re-identification device, as shown in Figure 6 The device comprises:

[0165] The acquisition module 601 is configured to acquire a to-be-identified image, wherein the to-be-identified image comprises a to-be-identified object.

[0166] The first determination module 602 is configured to determine a first quality score of a body region of the to-be-identified object in the to-be-identified image, and determine a second quality score of a face region of the to-be-identified object in the to-be-identified image.

[0167] The second determination module 603 is configured to determine a first similarity between the body region of the to-be-identified object and a body region of a target object, and determine a second similarity between the face region of the to-be-identified object and a face region of the target object.

[0168] The third determination module 604 is configured to determine a matching score of the to-be-identified object and the target object according to the first quality score, the second quality score, the first similarity and the second similarity.

[0169] The fourth determination module 605 is configured to determine whether the to-be-identified object and the target object are the same object according to the matching score.

[0170] In an embodiment of the present application, the first determination module 602 comprises:

[0171] The first determination sub-module is configured to detect the to-be-identified image to obtain the body region of the to-be-identified object in the to-be-identified image, and perform quality evaluation on the body region of the to-be-identified object to obtain the first quality score of the body region of the to-be-identified object.

[0172] The detection sub-module is configured to detect the body region of the to-be-identified object.

[0173] The second determination sub-module is configured to, if the face region of the to-be-identified object is detected, perform quality evaluation on the face region of the to-be-identified object to obtain the second quality score of the face region of the to-be-identified object.

[0174] The third determining sub-module is configured to determine that the second quality score of the face region of the to-be-identified object is 0 if the face region of the to-be-identified object is not detected.

[0175] In one embodiment of the present application, the third determining module 604 includes:

[0176] The first normalizing sub-module is configured to perform normalization processing on the first quality score and the second quality score respectively to obtain a first normalized score and a second normalized score.

[0177] The fourth determining sub-module is configured to determine a matching score of the to-be-identified object and the target object according to the first normalized score, the second normalized score, the first similarity and the second similarity.

[0178] In one embodiment of the present application, the third determining module 604 includes:

[0179] The fifth determining sub-module is configured to determine a first weight of the body region of the to-be-identified object and a second weight of the face region of the to-be-identified object according to the first quality score and the second quality score, wherein the first quality score is positively correlated with the first weight, and the second quality score is positively correlated with the second weight.

[0180] The sixth determining sub-module is configured to perform weighted summation on the first similarity and the second similarity by using the first weight and the second weight to obtain the matching score of the to-be-identified object and the target object.

[0181] In one embodiment of the present application, the fifth determining sub-module can be specifically configured to:

[0182] If the first quality score is higher than or equal to the preset body score threshold, and the second quality score is higher than or equal to the preset face score threshold, then the first weight of the body region of the to-be-identified object and the second weight of the face region of the to-be-identified object are determined as the ratios of the first quality score and the second quality score to a score sum value, and the score sum value is the sum of the first quality score and the second quality score.

[0183] If the first quality score is lower than the preset body score threshold, and the second quality score is higher than or equal to the preset face score threshold, then the first quality score is updated to 0, and the first weight and the second weight are determined as the ratios of the updated first quality score and the second quality score to the score sum value.

[0184] If the first quality score is higher than or equal to the preset body score threshold, and the second quality score is lower than the preset face score threshold, then the second quality score is updated to 0, and the first weight and the second weight are determined as the ratios of the first quality score and the updated second quality score to the score sum value.

[0185] In one embodiment of the present application, the fifth determining sub-module can be further configured to:

[0186] If the first quality score is lower than the preset body score threshold and the second quality score is lower than the preset face score threshold, it is determined that the to-be-identified object and the target object are not the same object.

[0187] In an embodiment of the present application, the first determination module 602 can be specifically used for:

[0188] performing weighted average processing on scores of a plurality of preset quality factors of the body region of the to-be-identified object in the to-be-identified image to obtain a first quality score, and performing weighted average processing on scores of a plurality of preset quality factors of the face region of the to-be-identified object in the to-be-identified image to obtain a second quality score; or

[0189] inputting the body region and the face region of the to-be-identified object in the to-be-identified image into a preset quality evaluation model to obtain the first quality score and the second quality score.

[0190] In the technical scheme provided by the embodiments of the present application, the first quality score of the body region of the to-be-identified object, the first similarity between the body region of the to-be-identified object and the body region of the target object, the second quality score of the face region of the to-be-identified object, and the second similarity between the face region of the to-be-identified object and the face region of the target object are used to determine whether the to-be-identified object and the target object are the same object, thereby completing the re-identification of the target object. In the case of comprehensively considering the information of the body region and the information of the face region, the false positives and false negatives of the object re-identification caused by changes in the information of the body region or poor image quality can be effectively avoided, and the accuracy of the object re-identification is improved.

[0191] In addition, face recognition is not sensitive to complex scenes such as object dressing, and changes in hairstyle or accessories, and has good robustness to complex scenes such as large poses, low image quality, strong / weak attention, and good generalization performance in infrared scenes. Therefore, the technical scheme provided by the embodiments of the present application uses the second quality score and the second similarity obtained by face recognition, in combination with the first quality score and the first similarity obtained by body recognition, to well cope with object re-identification in more complex scenes, greatly improve the generalization performance of object re-identification, and further improve the accuracy of object re-identification.

[0192] Corresponding to the above object re-identification method, the embodiments of the present application also provide an electronic device, as shown in the figure, comprising a processor 701 and a machine readable storage medium 702, the machine readable storage medium 702 stores machine executable instructions capable of being executed by the processor 701, and the processor 701 is prompted by the machine executable instructions to implement any of the steps of the above object re-identification method. Figure 7

[0193] ​The machine readable storage medium can include a random access memory (RAM), and can also include a non-volatile memory (NVM), such as at least one disk memory. Optionally, the machine readable storage medium can also be at least one storage device located away from the aforementioned processor.

[0194] The processor can be a general purpose processor, including a central processing unit (CPU), a network processor (NP), etc.; can also be a digital signal processor (DSP), an application specific integrated circuit (ASIC), a field-programmable gate array (FPGA) or other programmable logic device, a discrete gate or transistor logic device, a discrete hardware component.

[0195] Corresponding to the object recognition method described above, in another embodiment provided in the present application, an object recognition system is also provided, as shown in Figure 8 The object recognition system includes a camera 801 and an electronic device 802.

[0196] The camera 801 is configured to collect a to-be-identified image, and the to-be-identified image includes a to-be-identified object.

[0197] The electronic device 802 is configured to determine a first quality score of a body region of the to-be-identified object in the to-be-identified image, and determine a second quality score of a face region of the to-be-identified object in the to-be-identified image; determine a first similarity between the body region of the to-be-identified object and a body region of a target object, and determine a second similarity between the face region of the to-be-identified object and a face region of the target object; determine a matching score of the to-be-identified object and the target object according to the first quality score, the second quality score, the first similarity and the second similarity; and determine whether the to-be-identified object and the target object are the same object according to the matching score.

[0198] In the technical scheme provided by the embodiments of the present application, the first quality score of the body region of the to-be-identified object, the first similarity between the body region of the to-be-identified object and the body region of the target object, the second quality score of the face region of the to-be-identified object, and the second similarity between the face region of the to-be-identified object and the face region of the target object are used to determine whether the to-be-identified object and the target object are the same object, thereby completing object re-identification. In the case of comprehensively considering the information of the body region and the information of the face region, the false positives and false negatives of object re-identification caused by changes in the information of the body region or poor image quality can be effectively avoided, and the accuracy of object re-identification is improved.

[0199] In addition, face recognition is not sensitive to complex scenes such as object dressing scenes and scenes in which hairstyles or external decorations change, and the robustness of face recognition to complex scenes such as large poses, low image quality, strong / weak attention, and the like is good, and the generalization performance in infrared scenes is good. Therefore, the technical scheme provided by the embodiments of the present application uses the second quality score and the second similarity obtained by face recognition in combination with the first quality score and the first similarity obtained by body recognition, which can well cope with object re-identification in more complex scenes, greatly improves the generalization performance of object re-identification, and further improves the accuracy of object re-identification.

[0200] Corresponding to the above object re-identification method, in another embodiment provided by the present application, a computer readable storage medium is also provided, and the computer readable storage medium stores a computer program product. The computer program product is executed by a processor to implement any of the above object re-identification method steps.

[0201] Corresponding to the above object re-identification method, in another embodiment provided by the present application, a computer program product containing instructions is also provided, which, when running on a computer, causes the computer to execute any of the object re-identification methods in the above embodiments.

[0202] In the embodiments described above, all or some of the steps can be implemented by software, hardware or firmware, or any combination thereof. When implemented by software, all or some of the steps can be implemented in the form of one or more computer programs. The computer program can be stored in any computer readable medium, and loaded into the computer system for execution. The computer readable medium includes: a computer storage medium and a computer communication medium. The computer storage medium includes: volatile media (such as random access memory (RAM) and others) and non-volatile media (such as read-only memory (ROM), floppy disks, CD-ROMs, optical disks, hard disks, etc.). The computer communication medium includes: computer networks and other media.

[0203] It should be noted that, in this document, the terms such as first and second are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. In addition, the terms "comprise", "comprise" or any other variant thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or other elements inherent to such a process, method, article or device. Without more limitations, the element defined by the statement "comprises a" does not exclude the presence of additional identical elements in the process, method, article or device that includes the element.

[0204] Each of the embodiments in the specification is described in a related manner, and the same or similar parts between each of the embodiments can be referred to each other. Each of the embodiments focuses on the difference from other embodiments. In particular, for the device, electronic device, computer readable storage medium and computer program product embodiments, since they are basically similar to the method embodiments, the description is relatively simple, and the relevant parts can be referred to the part of the method embodiment.

[0205] The above merely provides the preferred embodiment of the present application, and not intended to limit the protection scope of the present application. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall be included in the protection scope of the present application.

Claims

1. A method for object re-identification, characterized in that, The method comprises: acquiring an image to be identified, the image to be identified comprising an object to be identified; determining a first quality score of a body region of the object to be identified in the image to be identified, and determining a second quality score of a face region of the object to be identified in the image to be identified; determining a first similarity between the body region of the object to be identified and a body region of a target object, and determining a second similarity between the face region of the object to be identified and a face region of the target object; performing similarity fusion on the first similarity and the second similarity by using the first quality score and the second quality score, to obtain a matching score of the object to be identified and the target object; the step of performing similarity fusion on the first similarity and the second similarity by using the first quality score and the second quality score, to obtain a matching score of the object to be identified and the target object, comprises: if the first quality score is higher than or equal to a preset body score threshold, and the second quality score is higher than or equal to a preset face score threshold, then determining a ratio of the first quality score and the second quality score to a score sum value, to obtain a first weight of the body region of the object to be identified and a second weight of the face region of the object to be identified, the score sum value being a sum of the first quality score and the second quality score; if the first quality score is lower than the preset body score threshold, and the second quality score is higher than or equal to the preset face score threshold, then updating the first quality score to 0, determining a ratio of the updated first quality score and the second quality score to the score sum value, to obtain the first weight and the second weight; if the first quality score is higher than or equal to the preset body score threshold, and the second quality score is lower than the preset face score threshold, then updating the second quality score to 0, determining a ratio of the first quality score and the updated second quality score to the score sum value, to obtain the first weight and the second weight; wherein the first quality score is positively correlated with the first weight, and the second quality score is positively correlated with the second weight; performing weighted summation on the first similarity and the second similarity by using the first weight and the second weight, to obtain the matching score of the object to be identified and the target object; determining whether the object to be identified and the target object are the same object according to the matching score, to complete re-identification of the target object.

2. The method of claim 1, wherein, the step of determining a first quality score of a body region of the object to be identified in the image to be identified, and determining a second quality score of a face region of the object to be identified in the image to be identified, comprises: detecting the image to be identified to obtain the body region of the object to be identified in the image to be identified, and performing quality evaluation on the body region of the object to be identified to obtain the first quality score of the body region of the object to be identified; detecting the body region of the object to be identified; If the face region of the to-be-identified object is detected, quality evaluation is performed on the face region of the to-be-identified object to obtain a second quality score of the face region of the to-be-identified object. If the face region of the to-be-identified object is not detected, it is determined that the second quality score of the face region of the to-be-identified object is 0.

3. The method of claim 1, wherein, The step of performing similarity fusion on the first similarity and the second similarity by using the first quality score and the second quality score to obtain a matching score of the to-be-identified object and the target object comprises: The first quality score and the second quality score are normalized respectively to obtain a first normalized score and a second normalized score. The matching score of the to-be-identified object and the target object is determined according to the first normalized score, the second normalized score, the first similarity and the second similarity.

4. The method of claim 1, wherein, The method further comprises: If the first quality score is lower than the preset body score threshold and the second quality score is lower than the preset face score threshold, it is determined that the to-be-identified object and the target object are not the same object.

5. The method of claim 1, wherein, The step of determining the first quality score of the body region of the to-be-identified object in the to-be-identified image and determining the second quality score of the face region of the to-be-identified object in the to-be-identified image comprises: The scores of a plurality of preset quality factors of the body region of the to-be-identified object in the to-be-identified image are weighted and averaged to obtain the first quality score, and the scores of the plurality of preset quality factors of the face region of the to-be-identified object in the to-be-identified image are weighted and averaged to obtain the second quality score; or The body region and the face region of the to-be-identified object in the to-be-identified image are respectively input into a preset quality evaluation model to obtain the first quality score and the second quality score.

6. An object re-identification apparatus characterized by comprising: The device comprises: An acquisition module configured to acquire a to-be-identified image, the to-be-identified image comprising a to-be-identified object; A first determination module configured to determine a first quality score of a body region of the to-be-identified object in the to-be-identified image and determine a second quality score of a face region of the to-be-identified object in the to-be-identified image; A second determination module configured to determine a first similarity between the body region of the to-be-identified object and a body region of a target object and determine a second similarity between the face region of the to-be-identified object and a face region of the target object; A third determination module configured to perform similarity fusion on the first similarity and the second similarity by using the first quality score and the second quality score to obtain a matching score of the to-be-identified object and the target object; A fourth determination module configured to determine whether the to-be-identified object and the target object are the same object according to the matching score, and complete re-identification of the target object; A fourth determination module configured to determine whether the to-be-identified object and the target object are the same object according to the matching score, and complete re-identification of the target object; The third determining module is specifically configured to: if the first quality score is higher than or equal to a preset body score threshold and the second quality score is higher than or equal to a preset face score threshold, determining a ratio of the first quality score and the second quality score to a score sum value, obtaining a first weight of the body region of the to-be-identified object and a second weight of the face region of the to-be-identified object, the score sum value being a sum of the first quality score and the second quality score; if the first quality score is lower than the preset body score threshold and the second quality score is higher than or equal to the preset face score threshold, updating the first quality score to 0, determining a ratio of the updated first quality score and the second quality score to the score sum value, and obtaining the first weight and the second weight; if the first quality score is higher than or equal to the preset body score threshold and the second quality score is lower than the preset face score threshold, updating the second quality score to 0, determining a ratio of the first quality score and the updated second quality score to the score sum value, and obtaining the first weight and the second weight; wherein the first quality score is positively correlated with the first weight, and the second quality score is positively correlated with the second weight; performing weighted summation on the first similarity and the second similarity by using the first weight and the second weight, to obtain a matching score of the to-be-identified object and the target object.

7. An electronic device, comprising: The system comprises a processor and a machine readable storage medium, the machine readable storage medium stores machine executable instructions which can be executed by the processor, and the processor is prompted by the machine executable instructions to implement the method steps of any one of claims 1-5.

8. A system for object re-identification, the system comprising: The system comprises a camera and an electronic device; The camera is configured to capture a to-be-identified image, the to-be-identified image comprising a to-be-identified object; The electronic device is configured to determine a first quality score of a body region of the to-be-identified object in the to-be-identified image, and determine a second quality score of a face region of the to-be-identified object in the to-be-identified image; determine a first similarity between the body region of the to-be-identified object and a body region of a target object, and determine a second similarity between the face region of the to-be-identified object and a face region of the target object; perform similarity fusion on the first similarity and the second similarity by using the first quality score and the second quality score, to obtain a matching score of the to-be-identified object and the target object; and determine whether the to-be-identified object and the target object are the same object according to the matching score, to complete re-identification of the target object; The electronic device is specifically configured to: if the first quality score is higher than or equal to a preset body score threshold, and the second quality score is higher than or equal to a preset face score threshold, then determine a ratio of the first quality score and the second quality score to a score sum value, to obtain a first weight of a body region and a second weight of a face region of the to-be-identified object, the score sum value being a sum of the first quality score and the second quality score; if the first quality score is lower than the preset body score threshold, and the second quality score is higher than or equal to the preset face score threshold, then update the first quality score to 0, and determine a ratio of the updated first quality score and the second quality score to the score sum value, to obtain the first weight and the second weight; if the first quality score is higher than or equal to the preset body score threshold, and the second quality score is lower than the preset face score threshold, then update the second quality score to 0, and determine a ratio of the first quality score and the updated second quality score to the score sum value, to obtain the first weight and the second weight; wherein the first quality score is positively correlated with the first weight, and the second quality score is positively correlated with the second weight; and performing weighted summation on the first similarity and the second similarity by using the first weight and the second weight, to obtain a matching score of the to-be-identified object and the target object.

Images