Target object identification method and apparatus, and device and storage medium

Abstract

Provided in the present disclosure are a target object identification method and apparatus, and a device and a storage medium. The method comprises: acquiring a target video, and identifying respective motion trajectories of one or more target objects in the target video; extracting a plurality of attribute features of the target objects from the motion trajectories of the target objects, and constructing a heterogeneous graph between the attribute features of the target objects and pre-stored base library features, wherein the attribute features and the base library features serve as nodes in the heterogeneous graph, and there are initial connection relationships between some of the nodes; and updating the initial connection relationships in the heterogeneous graph, determining from the base library features matching results of the attribute features of the target objects on the basis of the updated connection relationships between the nodes, and identifying the target objects on the basis of the matching results.

Description

Method, device, equipment and storage medium for identifying a target object

[0001] This application claims priority to the Chinese patent application filed with the State Intellectual Property Office of China on September 28, 2023, with application number 202311280241.1 and invention name “A method, device, equipment and storage medium for identifying a target object”, the entire contents of which are incorporated by reference into this application. Technical Field

[0002] The present disclosure relates to the field of image processing technology, and in particular to a method, apparatus, device, and storage medium for identifying a target object. Background Art

[0003] In some practical application scenarios, it is often necessary to identify target objects in images or videos. For example, currently, target objects can be identified using one-stage or two-stage object detection algorithms.

[0004] Due to the variability of actual application scenarios and the uneven quality of video frames, the recognition accuracy of target objects is not high. In view of this, a more accurate target object recognition method is currently needed.

[0005] Summary of the Invention

[0006] In view of this, one or more embodiments of the present disclosure provide a method, apparatus, device, and storage medium for identifying a target object.

[0007] On the one hand, the present disclosure provides a method for identifying a target object, the method comprising: acquiring a target video and identifying the motion trajectory of one or more target objects in the target video; extracting multiple attribute features of the target object from the motion trajectory of the target object, and constructing a heterogeneous graph between the attribute features of the target object and pre-stored underlying database features; wherein the attribute features and the underlying database features serve as nodes in the heterogeneous graph, and some nodes have initial connection relationships; updating the initial connection relationships in the heterogeneous graph, and based on the updated connection relationships between the nodes, determining matching results of each attribute feature of the target object in the underlying database features, and identifying the target object based on the matching results.

[0008] On the other hand, the present disclosure also provides a target object identification device, which includes: a trajectory identification unit, used to obtain a target video and identify the motion trajectory of one or more target objects in the target video; a heterogeneous graph construction unit, used to extract multiple attribute features of the target object from the motion trajectory of the target object, and construct a heterogeneous graph between the attribute features of the target object and pre-stored base library features; wherein the attribute features and the base library features serve as nodes in the heterogeneous graph, and some nodes have initial connection relationships; an object identification unit, used to update the initial connection relationship in the heterogeneous graph, and based on the updated connection relationship between the nodes, determine the matching results of each attribute feature of the target object in the base library features, and identify the target object based on the matching results.

[0009] On the other hand, the present disclosure further provides an electronic device, which includes a memory and a processor, wherein the memory is used to store a computer program, and when the computer program is executed by the processor, the above-mentioned target object recognition method is implemented.

[0010] On the other hand, the present disclosure further provides a computer-readable storage medium, which is used to store a computer program. When the computer program is executed by a processor, the above-mentioned target object recognition method is implemented.

[0011] The technical solution provided by one or more embodiments of the present disclosure utilizes a heterogeneous graph to establish an association between the attribute features of a target object and pre-stored underlying features. The attribute features of the target object are extracted from its motion trajectory and can describe the target object from multiple perspectives. The underlying features can serve as standard features of the target object, and the heterogeneous graph can be used to characterize the initial matching relationship between the attribute features and the underlying features.

[0012] Subsequently, by updating the initial connection relationship between each node in the heterogeneous graph, the matching relationship between the attribute features and the underlying database features can be accurately obtained, and the target object can be identified based on the matching relationship. BRIEF DESCRIPTION OF THE DRAWINGS

[0013] The features and advantages of the various embodiments of the present disclosure will be more clearly understood by referring to the accompanying drawings, which are schematic and should not be construed as limiting the present disclosure in any way. In the accompanying drawings:

[0014] FIG1 is a schematic diagram showing steps of a method for identifying a target object in one embodiment of the present disclosure;

[0015] FIG2 shows a schematic diagram of a heterogeneous graph in one embodiment of the present disclosure;

[0016] FIG3 shows a schematic diagram of an updated heterogeneous graph in one embodiment of the present disclosure;

[0017] FIG4 shows a target object identification flow chart in one embodiment of the present disclosure;

[0018] FIG5 shows a schematic diagram of functional modules of a target object recognition device according to an embodiment of the present disclosure; and

[0019] FIG6 shows a schematic structural diagram of an electronic device in one embodiment of the present disclosure. DETAILED DESCRIPTION

[0020] To make the purpose, technical solutions, and advantages of the embodiments of the present disclosure more clear, the technical solutions in the embodiments of the present disclosure will be clearly and completely described below in conjunction with the drawings in the embodiments of the present disclosure. Obviously, the described embodiments are part of the embodiments of the present disclosure, not all of the embodiments. Based on the embodiments of the present disclosure, all other embodiments obtained by those skilled in the art without making creative efforts shall fall within the scope of protection of the present disclosure.

[0021] In some scenarios where target object recognition is required, low recognition accuracy can occur due to factors such as the target's rapid movement or similar clothing. For example, at a ski resort, visitors often move at high speeds, resulting in blurred images of their figures. Recognizing visitors based on blurred images can result in low recognition accuracy. Furthermore, visitors often change into their ski suits after entering a ski resort, and since ski suits often have similar styles, identifying visitors solely based on their style can also result in low recognition accuracy.

[0022] In light of this, the present disclosure provides a target object recognition method that generates multiple different types of attribute features related to the target object. These attribute features can comprehensively describe the target object from different perspectives, providing a data foundation for improving recognition accuracy. Subsequently, a heterogeneous graph can be used to accurately match the attribute features with the underlying database features, thereby improving target object recognition accuracy.

[0023] Referring to FIG. 1 , a target object identification method provided by one embodiment of the present disclosure may include the following steps.

[0024] S1: Acquire a target video and identify the motion trajectory of one or more target objects in the target video.

[0025] In this embodiment, the target object may be the user to be identified. It is understood that before using the technical solutions disclosed in each embodiment of this disclosure, the user should be informed of the type, scope of use, and usage scenarios of the personal information involved in this disclosure in an appropriate manner in accordance with relevant laws and regulations and obtain the user's authorization.

[0026] For example, before implementing the technical solutions of various embodiments of the present disclosure, a prompt message may be sent to the user to clearly inform the user that the requested operation will require the acquisition and use of the user's personal information. This allows the user to independently choose whether to provide personal information to the electronic device, application, server, storage medium, or other software or hardware that performs the operations of the technical solutions of the present disclosure, based on the prompt message.

[0027] As an optional but non-limiting implementation, the prompt information can be sent to the user in the form of a pop-up window, in which the prompt information can be presented in text form. In addition, the pop-up window can also carry a selection control for the user to select "agree" or "disagree".

[0028] It is understandable that the above notification and user authorization process are merely illustrative and do not limit the implementation of the present disclosure. Other methods that comply with relevant laws and regulations may also be applied to the implementation of the present disclosure.

[0029] It is understandable that the data involved in this technical solution (including but not limited to the data itself, the acquisition or use of the data) must comply with the requirements of relevant laws, regulations and relevant provisions.

[0030] It is understandable that in the specific implementation of this application, when object identification, video data and other related data are involved, when the above embodiments of this application are applied to specific products or technologies, user permission or consent is required, and the collection, use and processing of relevant data must comply with relevant laws, regulations and standards of relevant countries and regions.

[0031] In this embodiment, through the camera deployed in the offline scene, when the target object is active in the offline scene, the target video containing the target object can be collected. Of course, the subsequent processing of the target video to identify the target object is carried out after obtaining the authorization of the target object. In practical applications, the target video can be analyzed frame by frame by the object detection algorithm. For any video frame in the target video, after processing with the object detection algorithm, the position information of one or more bounding boxes corresponding to the video frame can be output. Among them, each bounding box can represent a human body in the video frame, and the human body in the video frame can be identified and located through the position information of the bounding box.

[0032] Furthermore, after the object detection algorithm has been applied, each bounding box representing a human body will have its own object identifier, which can be, for example, a globally unique number. After object detection is performed on different frames of the target video, the aforementioned bounding box location information and object identifier can be output for each frame. Bounding boxes with the same object identifier can represent the same human body. Thus, after performing frame-by-frame detection on the target video, by identifying bounding boxes with the same object identifier, the motion trajectory of the same human body can be tracked across different video frames.

[0033] In this embodiment, the target object can be any human body to be identified in the target video. In the target video, the target object may only appear in a portion of the video frames and be identified by the object detection algorithm. In this way, by performing frame-by-frame object detection on the target video, the target object's motion trajectory can be determined in the target video.

[0034] It should be noted that, for different target objects, different motion trajectories can be obtained by the above method. Different motion trajectories can correspond to different target video frames in the target video. For example, for target object A, the motion trajectory can be represented by the target video frames from the 5th to the 30th second in the target video; while for target object B, the motion trajectory can be represented by the target video frames from the 1st to the 20th second, and from the 40th to the 50th second in the target video. In other words, for different target objects, the target video frames representing their motion trajectories can also be different.

[0035] S3: Extract multiple attribute features of the target object from the motion trajectory of the target object, and construct a heterogeneous graph between the attribute features of the target object and pre-stored base database features; wherein the attribute features and the base database features serve as nodes in the heterogeneous graph, and some nodes have initial connection relationships.

[0036] In this embodiment, after identifying the target video frame representing the motion trajectory of the target object from the target video, more detailed feature recognition can be performed on the target object in the target video frame. Taking one of the target objects as an example, different types of feature sequences related to the target object can be extracted from the motion trajectory of the target object from different angles. For example, in a skiing scene, the above-mentioned different types of feature sequences can include at least one of a head portrait feature sequence, a human body feature sequence, and a clothing icon feature sequence. The number of feature sequences can be flexibly determined based on the requirements of object recognition accuracy and data processing efficiency in the actual application scenario.

[0037] For example, in a specific application example, the target video frames of the target object can be processed frame by frame, and the target object's portrait features, body features, and clothing icon features can be extracted from each target video frame. By summarizing the different types of features extracted from each target video frame, three different types of feature sequences can be obtained (respectively, portrait feature sequence, body feature sequence, and clothing icon feature sequence).

[0038] In one embodiment, taking the feature sequence of any target type as an example, type features related to the target object and conforming to the target type can be identified from each target video frame. In practical applications, a feature extraction network can be used to extract the above-mentioned type features from the target video frame. Specifically, the feature extraction network can be flexibly selected from a variety of different network models according to the needs of the actual application scenario. The above-mentioned various network models may include, for example, a VGG model, an Inception model, a ResNet model, etc. Before the target video frame is input into the feature extraction network, the target video frame can be cropped to different degrees based on the different target types.

[0039] For example, when the target type is a head image, the target object's head region can be cropped from the target video frame and the cropped region image can be input into the feature extraction network, ensuring that the features output by the feature extraction network are head image features. Similarly, if the target type is a human body, the target object's bounding box can be directly cropped from the target video frame and the cropped region image can be input into the feature extraction network.

[0040] In this embodiment, after the corresponding type features are identified from each target video frame, the feature sequence composed of the identified type features can be used as the feature sequence of the target type. For different target types, respective feature sequences can be generated in a similar manner. It can be seen that a feature sequence can include multiple type features of the same type. For example, in the avatar feature sequence, the avatar features of the target object identified from each target video frame can be included. For another example, in the clothing icon feature sequence, the clothing icon features of the target object identified from each target video frame can be included.

[0041] In one embodiment, when extracting features from a target video frame and generating different types of type features, quality information of the type features can also be generated. The quality information of the type features can characterize the quality of the currently extracted type features. In practical applications, the target object in the target video frame may be blocked or move too fast, resulting in the type features of the target object extracted from the target video frame being not ideal, and the quality information can quantitatively evaluate the quality of each type feature. Different models can be used to generate quality information of type features for different type features.

[0042] For example, for avatar features, the megaface model can be used to output the quality information of the avatar features. For human body features, the human posture recognition (Pose Estimation) algorithm can be used to identify the posture information corresponding to the human body features. By comparing the posture information with the standard posture, the quality information corresponding to the human body features can be generated. Among them, the closer the identified posture information is to the standard posture, the better the quality of the quality information representation of the human body features. For clothing icon features, the quality information can be represented by parameters such as image contrast and image brightness. In this way, corresponding quality information can be generated for different types of type features. In practical applications, by combining the feature extraction model and the quality information generation model, when generating different types of feature sequences, the quality information of each type of feature in the feature sequence can also be generated together.

[0043] In this embodiment, when a target subject enters an offline scene, image information of the target subject that complies with preset rules can be captured, with the target subject's authorization. For example, when a visitor enters a ski resort, a clear frontal photo can be taken under the guidance of staff. These preset rules can also be limited to information such as the target subject's attire and posture when capturing image information. For example, when capturing a frontal photo of a visitor, the visitor must change into ski clothing and appropriate ski accessories, and then be photographed in a specified posture.

[0044] In this embodiment, after collecting image information of the target object that meets the preset rules, the type features of each preset type related to the target object can be identified from the collected image information according to the feature extraction method in the aforementioned steps. For example, the head portrait features, body features, and clothing icon features of the target object can be extracted from the collected image information. Since the image information is collected under relatively ideal conditions, the type features of each preset type identified from the image information can be used as the base database features of the target object, and the base database features can accurately characterize the target object. At the same time, after collecting the base database features of the target object, the standard information of the target object can also be bound to these base database features. In this way, for the same target object, multiple base database features extracted based on the image information of the target object can be bound to the standard information of the target object. Among them, the standard information can represent the correct object identification of the target object, and the standard information can be a user identification entered after confirmation by the target object when collecting the image information.

[0045] Through the above approach, for any playable object active in an offline scene, image information that conforms to pre-set criteria can be obtained. Subsequently, the type features of each pre-set category associated with the playable object can be identified from the image information, and each of these identified type features can be used as the base database features of the playable object. Furthermore, the base database features of the playable object are bound to the standard information of the playable object.

[0046] In this embodiment, after identifying multiple sets of different types of feature sequences of a target object from a target video frame, a representative feature can be determined from each feature sequence. Generally speaking, the representative feature can be the feature with the highest quality among the feature sequences of the corresponding type. Subsequently, the object identity of the target object can be determined based on the representative features in each feature sequence.

[0047] In one embodiment, when determining the representative features of each feature sequence, taking any target feature sequence as an example, the quality information of each type feature in the target feature sequence can be counted. This quality information can usually be a numerical value that can characterize the quality. Based on the statistical quality information, the type feature with the best quality can be used as the representative feature of the target feature sequence. The quality can be determined by the size of the numerical value. The larger the numerical value, the better the quality. Of course, if there are multiple type features with the best quality in the same feature sequence, then one of the multiple type features with the best quality can be randomly selected as the representative feature of the feature sequence.

[0048] In this embodiment, by statistically analyzing the quality information of type features, the type features with the highest quality from each feature sequence can be selected as the representative features of the feature sequence. For a target object, the representative features determined from each set of feature sequences can serve as multiple attribute features of the target object. Thus, multiple different attribute features can correspond to different types. Taking a skiing scene as an example, the multiple attribute features extracted from the target object's motion trajectory can include the target object's portrait features, body features, and clothing icon features.

[0049] In this embodiment, after identifying multiple attribute features of the target object, each attribute feature can be matched with the corresponding type of underlying database features in turn, and based on the matching results, a heterogeneous graph between the attribute features of the target object and the pre-stored underlying database features can be constructed.

[0050] Specifically, a heterogeneous graph can include multiple nodes and connections between some of these nodes. The various attribute features of the target object and the pre-stored underlying database features can serve as nodes in the heterogeneous graph. Based on the matching relationships between the attribute features and the underlying database features, initial connections can be established between some of the nodes. These initial connections can be reflected in the heterogeneous graph as edges connecting the nodes.

[0051] In this embodiment, the initial connection relationship between some nodes in the heterogeneous graph can be divided into two cases. One case is that for the same target object, a first initial connection relationship can be established between nodes represented by multiple attribute features of the target object. As shown in Figure 2, assuming that a target object has three attribute features: avatar features, body features, and clothing icon features, then in the heterogeneous graph, the nodes represented by these three attribute features can have edges representing the first initial connection relationship. In Figure 2, the rectangle can represent the avatar feature, the circle can represent the body feature, and the triangle can represent the clothing icon feature. It should be noted that in order to avoid confusion of the lines in Figure 2, only part of the initial connection relationship is shown in Figure 2.

[0052] In a heterogeneous graph, in addition to the first initial connection relationship described above, for any target attribute feature among the various attribute features of the target object, a target base database feature of the same type as the target attribute feature can be determined from the pre-stored base database features, and a second initial connection relationship can be established between the node represented by the target attribute feature and the nodes represented by each of the target base database features. As shown in Figure 2, the pre-stored base database features can be multiple target objects, and these base database features can also be divided into base database features of avatar type, base database features of human body type, and base database features of clothing icon type. In this way, for attribute features and base database features of the same type, an edge representing the second initial connection relationship can be established between their respective nodes.

[0053] As can be seen from Figure 2, there can be multiple target objects. Thus, within each target object, a first initial connection relationship can be established between nodes represented by multiple attribute features. For the attribute features of the target object, a second initial connection relationship can be established between the attribute features of the same type and the base database features.

[0054] In this embodiment, in order to determine the object identifier of the target object based on the heterogeneous graph, an initial weight value can be set for each initial connection relationship. For the first initial connection relationship and the second initial connection relationship, the initial weight values ​​can be set based on different rules.

[0055] Specifically, the first initial connection relationship represents the connection relationship between the attribute features of the same target object. In view of this, a preset fixed value can be set as the initial weight value of the first initial connection relationship. In actual application, the preset fixed value can be 1, indicating that the correlation between the attribute features of the target object is the highest.

[0056] The initial weight value of the second initial connection relationship can be set based on the similarity between the attribute feature and the base database feature. Specifically, taking any target attribute feature among the various attribute features of the target object as an example, a target base database feature of the same type as the target attribute feature can be first determined from the pre-stored base database features. For example, assuming the target attribute feature is of the avatar type, during feature matching, the target attribute feature will only be matched with base database features representing avatar features, and will not be matched with base database features representing human features.

[0057] In practical applications, the matching process between attribute features and underlying database features can be performed by calculating similarity. Specifically, the similarity between the target attribute feature and each target underlying database feature of the same type can be calculated separately. Since both the target attribute feature and the target underlying database feature can be represented by vectors, the similarity between the target attribute feature and the target underlying database feature can be expressed by the Euclidean distance between the two vectors. The closer the Euclidean distance, the more similar the two features are.

[0058] In this embodiment, through the above method, for the target attribute features and target base database features that have a second initial connection relationship, the similarity between the target attribute features and the target base database features can be calculated, and the similarity can be set as the initial weight value of the second initial connection relationship.

[0059] As shown in Figure 2, the values ​​marked on the edges between nodes represent the initial weights mentioned above. It can be seen that for attribute features of the same target object, the initial weight of the edge can be set to 1; and for attribute features and underlying database features of the same type, the initial weight of the edge can be set to the similarity between the attribute features and the underlying database features.

[0060] Through the above method, a heterogeneous graph between attribute features and base database features can be established. In the heterogeneous graph, attribute features and base database features can be used as nodes respectively, and some nodes can have initial connection relationships, which are also bound to initial weight values.

[0061] S5: updating the initial connection relationship in the heterogeneous graph, and determining the matching results of the various attribute features of the target object in the base database features based on the updated connection relationship between the nodes, and determining the object identifier of the target object according to the matching results.

[0062] In this embodiment, the initial connection relationship between the nodes in the heterogeneous graph, and the initial weight value bound to the initial connection relationship, can preliminarily reflect the degree of association between different features. However, the initial connection relationship and initial weight value determined in the above manner may not be consistent with the degree of association between the actual features. For example, in Figure 2, the avatar feature in the first target object has an initial connection relationship with each base database feature of the same type, and after similarity calculation, three matching results with similarities of 0.8, 0.7, and 0.2 will be obtained. However, in actual applications, the avatar feature will usually only correspond to one base database feature, and will not correspond to multiple different base database features at the same time. In addition, the avatar feature will often not correspond to multiple base database features with high similarity.

[0063] Given this, in order to accurately characterize the degree of association between nodes through edge weights, the initial weights binding the initial connection relationships in the heterogeneous graph can be updated. Based on the updated weights, the initial connection relationships in the heterogeneous graph are re-altered so that for any attribute feature, there is at most one underlying feature with a connection relationship in the updated heterogeneous graph. In other words, based on the updated weights, some of the initial connection relationships that originally existed in the heterogeneous graph can be canceled, and the heterogeneous graph with these canceled initial connection relationships can be used as the updated heterogeneous graph.

[0064] In order to achieve the above-mentioned goals, in practical applications, the GNN (Graph Neural Networks) algorithm, or a variant algorithm of the GNN algorithm, such as the GraphSAGE (SAmple and aggreGatE) algorithm, can be used to update the heterogeneous graph. Specifically, the characteristics of each node in the heterogeneous graph (attribute characteristics or underlying database characteristics), the initial connection relationship between nodes, and the initial weight value bound to each initial connection relationship can be used as input data of the algorithm. When the GNN algorithm or a variant algorithm of GNN processes the input data, the characteristics of each node can be re-aggregated to redetermine the degree of association between the nodes. After processing by the GNN algorithm or a variant algorithm of GNN, the data of the re-aggregated heterogeneous graph can be output. Specifically, in the re-aggregated heterogeneous graph, the initial weight value can be updated according to the aggregation result.

[0065] In this embodiment, after the initial weight values ​​in the heterogeneous graph are updated, the connection relationship between the nodes in the heterogeneous graph can be further changed according to the updated weight values ​​to ensure that the same attribute feature has at most one base database feature with a connection relationship.

[0066] In practical applications, a weighted bipartite graph matching algorithm can be used to update the connection relationship between nodes in a heterogeneous graph. After processing with the weighted bipartite graph matching algorithm, a set of edges can be obtained, wherein the two vertices of each edge can be located in the set of attribute features and the set of base database features respectively, and no two edges have the same vertices. In this way, after processing with the weighted bipartite graph matching algorithm, some of the initial connection relationships that originally existed in the heterogeneous graph can be cancelled. For example, in Figure 2, there was an initial connection relationship between the avatar feature of the original first target object and the base database features that represent each avatar type, but after processing with the weighted bipartite graph matching algorithm, the avatar feature only retained the connection relationship with the largest weight value (as shown in Figure 3), and for the base database features connected to the avatar feature, there are no avatar features of other target objects connected. It should be noted that in order to avoid confusion of the lines in Figure 3, only part of the connection relationship is shown in Figure 3.

[0067] Through the above method, the initial weight value of the initial connection relationship in the heterogeneous graph can be updated, and based on the updated weight value, some of the initial connection relationships in the heterogeneous graph can be cancelled, so that for any attribute feature, in the updated heterogeneous graph, there is at most one base database feature with a connection relationship.

[0068] In this embodiment, after the heterogeneous graph is updated, the matching results of the various attribute features of the target object in the underlying database features can be determined based on the updated heterogeneous graph. Specifically, for any target attribute feature among the various attribute features of the target object, a target underlying database feature that has a connection relationship with the target attribute feature can be identified in the updated heterogeneous graph, and the target underlying database feature can be used as the underlying database feature that matches the target attribute feature.

[0069] Of course, in some scenarios, the target attribute features may not have any underlying database features with a connected relationship in the updated heterogeneous graph. For example, in Figure 3, the human features of the second target object do not have any underlying database features with a connected relationship. This indicates that the recognition results based on the target video cannot associate the attribute features with the pre-stored underlying database features. This may be because the corresponding underlying database features are not pre-stored for the target object. In other words, the visitor in the offline scene may not have initially entered the underlying database features, or the visitor does not have permission to match the underlying database features. Therefore, based on the captured target video, no matching underlying database features can be found. Alternatively, the attribute features identified from the target video may not be accurate enough. For example, if the target object is too blurry in the target video, the identified attribute features of the target object will be inaccurate, making it impossible to match them with the pre-stored underlying database features.

[0070] As can be seen from the above description, for each attribute feature of a target object, each attribute feature will only match at most one underlying feature. Therefore, when determining the object ID of the target object, we can first count whether each attribute feature of the target object has a matching underlying feature. If each representative feature has a matching underlying feature, we can then count the standard information of each matching underlying feature. Based on this count of standard information, we can then determine the object ID of the target object.

[0071] Refer to Figure 4. Specifically, if the statistically analyzed standard information corresponds to the same object identifier, then this same object identifier can be used as the object identifier of the target object. For example, after matching the three attribute features of the target object with the underlying database features, matching underlying database features O, P, and Q can be obtained. If the standard information bound to these three underlying database features O, P, and Q all points to user M, then user M can be used as the object identifier of the target object.

[0072] On the other hand, if the various standard information counted corresponds to different object identifiers, it means that the object identifiers corresponding to the various attribute features are inconsistent. At this time, a prompt message indicating that the object identifier cannot be confirmed can be generated. For example, after the three attribute features of the target object are matched with the base database features, matching base database features D, E, and F can be obtained. However, the standard information bound to base database features D and E both point to user K, while the standard information bound to base database feature F points to user L. This indicates that there is a contradiction in the object identification in the matching result, and the object identification of the target object cannot be determined. The object identification of the target object can be further confirmed through other video data later, but in the current round of data processing, the object identification of the target object cannot be confirmed.

[0073] Furthermore, in one embodiment, considering that not every attribute feature has a matching underlying database feature, if at least one of the target object's attributes does not have a matching underlying database feature, a prompt message indicating that the object's identity cannot be confirmed can be directly generated. This allows for a direct determination of the object's identity failure through a preliminary examination of the matching results, avoiding the subsequent statistical analysis of standard information and reducing data processing complexity.

[0074] The technical solution provided by one or more embodiments of the present disclosure utilizes a heterogeneous graph to establish an association between the attribute features of a target object and pre-stored underlying features. The attribute features of the target object are extracted from its motion trajectory and can describe the target object from multiple perspectives. The underlying features can serve as standard features of the target object, and the heterogeneous graph can be used to characterize the initial matching relationship between the attribute features and the underlying features.

[0075] Subsequently, by updating the initial connection relationship between each node in the heterogeneous graph, the matching relationship between the attribute features and the underlying database features can be accurately obtained. Based on this matching relationship, the target object can be identified.

[0076] As can be seen, one or more embodiments of the present disclosure can analyze the target object's motion trajectory when confirming its object identification, thereby avoiding the problem of inaccurate recognition results caused by poor quality of a single video frame. Furthermore, multiple attribute features can comprehensively describe the target object from multiple perspectives. Using a heterogeneous graph, these multiple attribute features can be accurately matched with pre-stored underlying features, thereby improving the accuracy of target object recognition.

[0077] Please refer to Figure 5. An embodiment of the present disclosure also provides a target object identification device, which includes: a trajectory identification unit 100, used to obtain a target video and identify the motion trajectory of one or more target objects in the target video; a heterogeneous graph construction unit 200, used to extract multiple attribute features of the target object from the motion trajectory of the target object, and construct a heterogeneous graph between the attribute features of the target object and pre-stored base library features; wherein the attribute features and the base library features serve as nodes in the heterogeneous graph, and some nodes have initial connection relationships; an object identification unit 300, used to update the initial connection relationship in the heterogeneous graph, and based on the updated connection relationship between the nodes, determine the matching results of each attribute feature of the target object in the base library features, and identify the target object based on the matching results.

[0078] The specific processing logic of each functional module can be found in the description of the aforementioned method implementation method, which will not be repeated here.

[0079] The various units described in the above embodiments can be implemented by computer chips or products with certain functions. A typical implementation device is a computer. Specifically, the computer can be, for example, a personal computer, a laptop computer, a cellular phone, a camera phone, a smartphone, a personal digital assistant, a media player, a navigation device, an email device, a game console, a tablet computer, a wearable device, or a combination of any of these devices.

[0080] For the convenience of description, the above devices are described as being divided into various units according to their functions. Of course, when implementing this application, the functions of each unit can be implemented in the same or multiple software and / or hardware.

[0081] Please refer to FIG6 . An embodiment of the present disclosure further provides an electronic device, which includes a memory and a processor. The memory is used to store a computer program. When the computer program is executed by the processor, the above-mentioned target object recognition method is implemented.

[0082] The present disclosure also provides a computer-readable storage medium, which is used to store a computer program. When the computer program is executed by a processor, the above-mentioned target object recognition method is implemented.

[0083] The processor may be a central processing unit (CPU). The processor may also be other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, or a combination of the above chips.

[0084] Memory, as a non-transitory computer-readable storage medium, can be used to store non-transitory software programs, non-transitory computer executable programs, and modules, such as the program instructions / modules corresponding to the methods in the embodiments of the present disclosure. The processor executes the non-transitory software programs, instructions, and modules stored in the memory to perform various processor functions and data processing, thereby implementing the methods in the aforementioned method embodiments.

[0085] The memory may include a program storage area and a data storage area, wherein the program storage area may store an operating system, an application required for at least one function; the data storage area may store data created by the processor, etc. In addition, the memory may include a high-speed random access memory, and may also include a non-transitory memory, such as at least one disk storage device, a flash memory device, or other non-transitory solid-state storage device. In some embodiments, the memory may optionally include a memory remotely located relative to the processor, and these remote memories may be connected to the processor via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0086] Those skilled in the art will appreciate that all or part of the processes in the above-mentioned embodiments can be implemented by instructing the relevant hardware through a computer program. The program can be stored in a computer-readable storage medium. When the program is executed, it can include the processes of the above-mentioned embodiments. The storage medium can be a magnetic disk, an optical disk, a read-only memory (ROM), a random access memory (RAM), a flash memory, a hard disk drive (HDD), or a solid-state drive (SSD). The storage medium can also include a combination of the above-mentioned types of memory.

[0087] The various embodiments in this specification are described in a progressive manner. Similar portions between the various embodiments can be referenced to each other, and each embodiment focuses on the differences from the other embodiments. In particular, the device, equipment, and storage medium embodiments are generally similar to the method embodiments, so their descriptions are relatively simplified. For relevant portions, refer to the descriptions of the method embodiments.

[0088] The foregoing is merely an embodiment of the present application and is not intended to limit the present application. For those skilled in the art, the present application may have various changes and variations. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present application should all be included within the scope of the claims of the present application.

[0089] Although the embodiments of the present disclosure have been described with reference to the accompanying drawings, those skilled in the art may make various modifications and variations without departing from the spirit and scope of the present disclosure, and such modifications and variations are all within the scope defined by the appended claims.

Claims

1. A method for identifying a target object, the method comprising: Acquire a target video, and identify the motion trajectory of one or more target objects in the target video; Extracting multiple attribute features of the target object from the motion trajectory of the target object, and constructing a heterogeneous graph between the attribute features of the target object and pre-stored base database features; wherein the attribute features and the base database features serve as nodes in the heterogeneous graph, and some nodes have initial connection relationships; The initial connection relationship in the heterogeneous graph is updated, and based on the connection relationship between the updated nodes, the matching results of the various attribute features of the target object in the base database features are determined, and the target object is identified according to the matching results.

2. The method according to claim 1, wherein extracting a plurality of attribute features of the target object from the motion trajectory of the target object comprises: In the motion trajectory of the target object, identifying multiple groups of different types of feature sequences related to the target object; Representative features of each group of feature sequences are determined respectively, and each of the representative features is used as a plurality of attribute features of the target object.

3. The method according to claim 2, wherein in the motion trajectory of the target object, identifying a plurality of groups of different types of feature sequences related to the target object comprises: For any target type among the different types, identifying a type feature related to the target object and conforming to the target type from the motion trajectory of the target object; The feature sequence formed by the identified features of each type is used as the feature sequence of the target type.

4. The method according to claim 2, wherein each group of the feature sequences contains a plurality of respective type features, and each of the type features has respective quality information; and determining respective representative features from each group of the feature sequences comprises: For any target feature sequence in the multiple groups of different types of feature sequences, counting quality information of each type of feature in the target feature sequence; Based on the statistically obtained quality information, the type feature with the best quality is used as the representative feature of the target feature sequence.

5. The method according to claim 1, wherein constructing a heterogeneous graph between the attribute features of the target object and the pre-stored base database features comprises: Establishing a first initial connection relationship between nodes represented by each attribute feature of the target object; For any target attribute feature among the attribute features of the target object, a target base database feature that belongs to the same type as the target attribute feature is determined in the pre-stored base database features, and a second initial connection relationship is established between the node represented by the target attribute feature and the nodes represented by each of the target base database features.

6. The method according to claim 5, further comprising: Setting a preset fixed value as an initial weight value of the first initial connection relationship; For the target attribute feature and the target base database feature that have a second initial connection relationship, the similarity between the target attribute feature and the target base database feature is calculated, and the similarity is set as an initial weight value of the second initial connection relationship.

7. The method according to claim 1, wherein updating the initial connection relationship in the heterogeneous graph comprises: The initial weight values ​​of the initial connection relationships in the heterogeneous graph are updated, and based on the updated weight values, some of the initial connection relationships in the heterogeneous graph are cancelled, and the heterogeneous graph with some of the initial connection relationships cancelled is used as the updated heterogeneous graph.

8. The method according to claim 1 or 7, wherein determining the matching result of each attribute feature of the target object in the base database feature comprises: For any target attribute feature among the attribute features of the target object, a target base database feature that is connected to the target attribute feature is identified in the updated heterogeneous graph, and the target base database feature is used as a base database feature that matches the target attribute feature.

9. The method according to claim 1, wherein according to the matching result, identifying the target object comprises: If each of the attribute features of the target object has a respective matching base database feature, statistically analyzing the standard information of each of the matching base database features; The object identifier of the target object is determined according to the statistically obtained standard information.

10. The method according to claim 9, further comprising: If at least one of the attribute features of the target object does not have a matching base database feature, a prompt message indicating that the object identifier cannot be confirmed is generated.

11. The method according to claim 9, wherein determining the object identification of the target object according to the statistically obtained standard information comprises: If each piece of statistically obtained standard information corresponds to the same object identifier, the same object identifier is used as the object identifier of the target object; If each piece of statistically obtained standard information corresponds to a different object identifier, a prompt message indicating that the object identifier cannot be confirmed is generated.

12. A device for identifying a target object, the device comprising: A trajectory recognition unit, used to acquire a target video and recognize the motion trajectory of one or more target objects in the target video; A heterogeneous graph construction unit is used to extract multiple attribute features of the target object from the motion trajectory of the target object, and construct a heterogeneous graph between the attribute features of the target object and pre-stored base database features; wherein the attribute features and the base database features are used as nodes in the heterogeneous graph, and some nodes have an initial connection relationship; The object identification unit is used to update the initial connection relationship in the heterogeneous graph, and based on the connection relationship between the updated nodes, determine the matching results of various attribute features of the target object in the base database features, and identify the target object according to the matching results.

13. An electronic device, comprising a memory and a processor, wherein the memory is used to store a computer program, and when the computer program is executed by the processor, the method according to any one of claims 1 to 11 is implemented.

14. A computer-readable storage medium, wherein the computer-readable storage medium is used to store a computer program, wherein when the computer program is executed by a processor, the method according to any one of claims 1 to 11 is implemented.

Images