Information processing method and device, electronic equipment and storage medium

Abstract

The present disclosure provides an information processing method and device, electronic equipment and storage medium, which can be applied to the fields of artificial intelligence technology and financial technology. The method comprises: determining a target region from a target video frame, wherein the target region comprises a target object; generating a location-sensitive score map about the target object according to the target region, the location-sensitive score map comprising object part features of the target object; analyzing the object part features to obtain a behavior recognition result about the target object; and generating behavior prompt information about the target object according to the behavior recognition result.

Description

TECHNICAL FIELD

[0001] The present disclosure relates to the technical field of artificial intelligence and the technical field of financial technology, in particular to an information processing method and device, electronic equipment, medium and program product. BACKGROUND

[0002] Action recognition technology is a technology of determining the behavior intention of a target object by analyzing the image of the target object. For example, video frames can be extracted from a video, and the behavior of the target object in the video frames is recognized. Thus, the behavior of the target object can be prompted.

[0003] However, in the process of implementing the concept of the present disclosure, the inventors found that the efficiency of prompting the behavior of the target object in the related art is low, and the accuracy of prompting the behavior of the target object is low. SUMMARY

[0004] In view of the above problems, the present disclosure provides an information processing method and device, electronic equipment, medium and program product.

[0005] According to a first aspect of the present disclosure, an information processing method is provided, comprising: determining a target region from a target video frame, wherein the target region comprises a target object; generating a location-sensitive score map about the target object according to the target region, the location-sensitive score map comprising object part features of the target object; analyzing the object part features to obtain a behavior recognition result about the target object; and generating behavior prompt information about the target object according to the behavior recognition result.

[0006] According to an embodiment of the present disclosure, the target region is determined from the target video frame, comprising: performing feature extraction on the target video frame to obtain a target feature map; and performing labeling on the target feature map according to target bounding box size information to obtain the target region, wherein the target bounding box size information is determined according to bounding box size information, the bounding box size information comprises first bounding box size information and second bounding box size information, the first bounding box size information is set by a user, and the second bounding box size information is set by a target model.

[0007] According to an embodiment of the present disclosure, the target annotation box size information is determined according to the annotation box size information by the following operation, comprising: according to the annotation box size information of the t th round, annotating the target feature map to obtain the annotation region of the t th round, wherein t is a positive integer less than T, and T is the iteration round; determining K first center regions and J first non-center regions from the annotation region of the t th round, wherein K and J are positive integers; determining distance information between the k th center region in the K first center regions and the j th non-center region in the J first non-center regions, wherein k is a positive integer less than K, and j is a positive integer less than J; determining a target non-center region from the J first non-center regions, which satisfies a first predetermined condition with the k th center region; determining the annotation box size information of the t+1 th round according to the target non-center region; determining difference information between the annotation box size information of the t th round and the annotation box size information of the t+1 th round; in the case that the difference information satisfies a second predetermined condition, determining the annotation box size information of the t+1 th round as the target annotation box size information.

[0008] According to an embodiment of the present disclosure, the distance information between the k th center region in the K first center regions and the j th non-center region in the J first non-center regions is determined, comprising: determining the intersection area between the k th center region and the j th non-center region; determining the union area between the k th center region and the j th non-center region; and determining the distance information between the k th center region and the j th non-center region according to the intersection area and the union area.

[0009] According to an embodiment of the present disclosure, the target non-center region is M, M is a positive integer greater than 1, and the size information of the target non-center region includes a height value and a width value; and the size information of the center region of the t+1 th round is determined according to the target non-center region, comprising: determining the height value of the center region of the t+1 th round according to the median of the height values of the M target non-center regions; determining the width value of the center region of the t+1 th round according to the median of the width values of the M target non-center regions; and determining the size information of the center region of the t+1 th round according to the height value of the center region of the t+1 th round and the width value of the center region of the t+1 th round.

[0010] According to an embodiment of the present disclosure, the position-sensitive score map about the target object is generated according to the target region, comprising: generating sub-region size information according to a predetermined parameter and the size information of the target region; dividing the target region according to the sub-region size information to obtain Q target sub-regions, wherein Q is a positive integer greater than 1; determining confidence information corresponding to the Q target sub-regions; and generating the position-sensitive score map according to the confidence information and the target region.

[0011] According to an embodiment of the present disclosure, the determining the confidence information corresponding to the Q target sub-regions comprises: performing pooling on the Q target sub-regions to obtain a pooled region; and processing the pooled region to obtain the confidence information.

[0012] According to an embodiment of the present disclosure, the target video frame is F, and F is a positive integer; the analyzing the object part feature to obtain the behavior recognition result of the target object comprises: determining pose information of the target object according to the object part feature corresponding to the F target video frames; and recognizing the pose information to obtain the behavior recognition result of the target object.

[0013] A second aspect of the present disclosure provides an information processing apparatus, comprising: a determination module configured to determine a target region from a target video frame, wherein the target region comprises a target object; a first generation module configured to generate a location sensitive score map of the target object according to the target region, wherein the location sensitive score map comprises an object part feature of the target object; an analysis module configured to analyze the object part feature to obtain a behavior recognition result of the target object; and a second generation module configured to generate behavior prompt information of the target object according to the behavior recognition result.

[0014] A third aspect of the present disclosure provides an electronic device, comprising: one or more processors; and a memory storing one or more programs for execution by the one or more processors, wherein the one or more programs, when executed by the one or more processors, cause the one or more processors to perform the above method.

[0015] A fourth aspect of the present disclosure further provides a computer-readable storage medium having stored thereon executable instructions that, when executed by a processor, cause the processor to perform the above method.

[0016] A fifth aspect of the present disclosure further provides a computer program product comprising a computer program that, when executed by a processor, implements the above method.

[0017] According to the information processing method, apparatus, electronic device, medium and program product provided by the present disclosure, the location sensitive score map of the target object is generated according to only the target region, and then the full amount of regions in the target video frame can be avoided to be recognized, and the recognition efficiency is improved. Moreover, the behavior of the target object is analyzed according to the object part feature in the location sensitive score map, and the overall feature of the target object is avoided to be analyzed, the recognition granularity is improved, and the accuracy of recognizing the behavior of the target object is improved. Therefore, by improving the recognition efficiency and the recognition accuracy, the efficiency and the accuracy of the generated behavior prompt information of the target object can be improved. BRIEF DESCRIPTION OF DRAWINGS

[0018] The foregoing contents, as well as other objects, features, and advantages of this disclosure, will become clearer from the following description of embodiments with reference to the accompanying drawings, in which:

[0019] Figure 1 This diagram illustrates an application scenario of the information processing method according to an embodiment of the present disclosure.

[0020] Figure 2 A flowchart illustrating an information processing method according to an embodiment of the present disclosure is shown schematically.

[0021] Figure 3 A schematic diagram illustrating a prompting method according to an embodiment of the present disclosure is shown;

[0022] Figure 4 A schematic diagram of the target area according to an embodiment of the present disclosure is shown;

[0023] Figure 5 The illustration shows a schematic diagram of obtaining a target feature map according to an embodiment of the present disclosure;

[0024] Figure 6 A schematic diagram of a target sub-region according to an embodiment of the present disclosure is shown;

[0025] Figure 7 A schematic diagram illustrating the generation of a location-sensitive score map according to an embodiment of the present disclosure is shown.

[0026] Figure 8 A schematic diagram illustrating the acquisition of confidence information according to an embodiment of the present disclosure is shown.

[0027] Figure 9 A schematic block diagram of an information processing apparatus according to embodiments of the present disclosure is shown; and

[0028] Figure 10 A block diagram schematically illustrates an electronic device suitable for implementing an information processing method according to an embodiment of the present disclosure. Detailed Implementation

[0029] The embodiments of the present disclosure will now be described with reference to the accompanying drawings. However, it should be understood that these descriptions are exemplary only and are not intended to limit the scope of the disclosure. In the following detailed description, numerous specific details are set forth to provide a thorough understanding of the embodiments of the present disclosure for ease of explanation. However, it will be apparent that one or more embodiments may be practiced without these specific details. Furthermore, descriptions of well-known structures and techniques are omitted in the following description to avoid unnecessarily obscuring the concepts of the present disclosure.

[0030] The terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting of the disclosure. As used herein, the terms "comprises", "comprising", "includes", "including" and the like are specifically intended to be open-ended and to mean that other features, steps, operations, and / or components can be added.

[0031] All terms used herein, including technical and scientific terms, have the meanings commonly understood by one of ordinary skill in the art unless otherwise defined. It should be noted that the terms used herein are to be interpreted as having a meaning that is consistent with the context of the specification, and should not be interpreted in an idealized or overly formal way.

[0032] In the case of using expressions similar to "at least one of A, B, and C, etc.", it should be generally interpreted as including at least one of the items enumerated, but not limited to the items enumerated (e.g., "a system having at least one of A, B, and C" should include a system having A alone, a system having B alone, a system having C alone, a system having both A and B, a system having both A and C, a system having both B and C, and / or a system having A, B, and C, etc.).

[0033] In the technical solutions of the present application, the user information (including but not limited to user personal information, user image information, user equipment information, such as location information, etc.) and data (including but not limited to data for analysis, stored data, displayed data, etc.) involved are information and data authorized by the user or authorized by all parties, and the collection, storage, use, processing, transmission, provision, disclosure and application of related data comply with relevant laws, regulations and standards of relevant countries and regions, necessary security measures are taken, public order and good customs are not violated, and corresponding operation portals are provided for users to choose authorization or refusal.

[0034] However, in the process of implementing the concept of the present application, the inventors have found that the related art has low efficiency in prompting the behavior of a target object, and low accuracy in prompting the behavior of the target object.

[0035] For example, in the case of a large number of target objects included in a video, the efficiency of prompting the behavior of each target object is low, and the accuracy of prompting the behavior of each target object is low.

[0036] Therefore, embodiments of the present disclosure provide an information processing method, including: determining a target region from a target video frame, wherein the target region includes a target object; generating a location-sensitive score map about the target object according to the target region, the location-sensitive score map including object part features of the target object; analyzing the object part features to obtain a behavior recognition result about the target object; and generating behavior prompt information about the target object according to the behavior recognition result.

[0037] Figure 1 An application scenario diagram of the information processing method according to an embodiment of the present disclosure is shown schematically.

[0038] As shown in the figure, Figure 1 The application scenario 100 according to this embodiment can include a first terminal device 101, a second terminal device 102, a third terminal device 103, a network 104, and a server 105. The network 104 is a medium for providing a communication link between the first terminal device 101, the second terminal device 102, the third terminal device 103, and the server 105. The network 104 can include various connection types, such as wired, wireless communication links, or optical fiber cables, etc.

[0039] A user can use at least one of the first terminal device 101, the second terminal device 102, and the third terminal device 103 to interact with the server 105 through the network 104 to receive or send messages, etc. Various communication client applications can be installed on the first terminal device 101, the second terminal device 102, and the third terminal device 103, such as shopping applications, web browser applications, search applications, instant messaging tools, email clients, social platform software, etc. (only as examples).

[0040] The first terminal device 101, the second terminal device 102, or the third terminal device 103 can be various electronic devices with a display screen and supporting web browsing, including but not limited to smartphones, tablet computers, laptop computers, desktop computers, etc.

[0041] The server 105 can be a server providing various services, such as a background management server supporting a website browsed by a user using the first terminal device 101, the second terminal device 102, or the third terminal device 103 (only as an example). The background management server can analyze and process received user requests and other data, and feed back the processing results (such as web pages, information, or data, etc. obtained or generated according to user requests) to the terminal device.

[0042] It should be noted that the information processing method provided by the embodiments of the present disclosure can generally be executed by the server 105. Accordingly, the information processing apparatus provided by the embodiments of the present disclosure can generally be arranged in the server 105. The information processing method provided by the embodiments of the present disclosure can also be executed by a server or a server cluster different from the server 105 and capable of communicating with the first terminal device 101, the second terminal device 102, the third terminal device 103, and / or the server 105. Accordingly, the information processing apparatus provided by the embodiments of the present disclosure can also be arranged in a server or a server cluster different from the server 105 and capable of communicating with the first terminal device 101, the second terminal device 102, the third terminal device 103, and / or the server 105.

[0043] It should be understood that Figure 1 The number of terminal devices, networks and servers in the above-mentioned scenario is only illustrative. Any number of terminal devices, networks and servers can be provided according to implementation needs.

[0044] The information processing method according to the embodiments of the present disclosure will be described in detail below based on the scenario described above. Figure 1 Figures 2-8 The information processing method of the embodiments will be described in detail.

[0045] Figure 2 A flowchart of an information processing method according to an embodiment of the present disclosure is schematically shown.

[0046] As shown in Figure 2 The information processing method of the embodiments includes operations S210-S240.

[0047] In operation S210, a target region is determined from a target video frame, wherein the target region includes a target object.

[0048] In operation S220, a position-sensitive score map about the target object is generated according to the target region, wherein the position-sensitive score map includes object part features of the target object.

[0049] In operation S230, the object part features are analyzed to obtain a behavior recognition result about the target object.

[0050] In operation S240, behavior prompt information about the target object is generated according to the behavior recognition result.

[0051] According to an embodiment of the present disclosure, the target video frame can be obtained from a collected video. The target video frame can include a target region. The target region can be a region including a target object. The target object can be a living being or the like in the target video frame.

[0052] According to an embodiment of the present disclosure, the position-sensitive score map can include object part features of the target object and confidence information corresponding to the object part features. The object part features can include features of a head of the target object, features of upper limbs of the target object, or features of lower limbs of the target object, etc.

[0053] The confidence information corresponding to the object part features can include class probabilities corresponding to the object part features.

[0054] ​For example, the object part features can include object part feature A, object part feature B, and object part feature C. The confidence information corresponding to the object part feature A can be "a probability of belonging to a head is 92%"; the confidence information corresponding to the object part feature B can be "a probability of belonging to an upper limb is 97%"; and the confidence information corresponding to the object part feature C can be "a probability of belonging to a lower limb is 90%".

[0055] According to an embodiment of the present disclosure, the behavior recognition result can include a first result and a second result. The first result can be used to represent that the behavior of the target object is abnormal, and the second result can be used to represent that the behavior of the target object is normal.

[0056] According to an embodiment of the present disclosure, the target model can be used to process the target video frame to obtain a target region including the target object. The target model can be obtained by training a region proposal network (RPN). The target model can be trained by using a detection classification probability method and a detection bounding box regression method. The target model can be trained by using a learning rate dynamic replacement training method. The number of sample images for a single training of the target model can be 128.

[0057] According to an embodiment of the present disclosure, the target region can be P, P can be a positive integer, and the confidence corresponding to each of the P target regions can be calculated. In addition, a location sensitive score map about the target object can be generated according to the target region and the confidence corresponding to each of the target regions.

[0058] According to an embodiment of the present disclosure, the target video frame can be O, and the location sensitive score map can correspond to O, O being a positive integer. The action recognition model can be used to process the O location sensitive score maps to analyze the object part features of the target object corresponding to each of the O location sensitive score maps, to obtain behavior information of the target object. The behavior information can include posture information of the target object, etc. The behavior recognition result about the target object can be obtained according to the behavior information.

[0059] According to an embodiment of the present disclosure, in a case where the behavior recognition result is the first result, it can be determined that the target object has an abnormal condition, and behavior prompt information about the target object can be generated to prompt the behavior of the target object. In a case where the behavior recognition result is the second result, it is determined that the target object has no abnormality. In a case where the behavior recognition result is the first result, the posture information can include a prone posture information or a supine posture information, etc. The abnormal condition can include a falling condition of the target object, etc. The falling condition can correspond to the prone posture information or the supine posture information.

[0060] According to an embodiment of the present disclosure, the behavior prompt information can be generated according to pre-stored location information and behavior information of the target object. After generating the behavior prompt information about the target object, the method can further include: sending the behavior prompt information to a rescue platform, so that the rescue platform executes a processing strategy for the target object according to the location information and the behavior information in the behavior prompt information. The processing strategy can include a rescue strategy, etc. In this way, by rescuing the target object, the life safety of the target object can be ensured. The pre-stored location information can be pre-stored by the target object.

[0061] According to an embodiment of the present disclosure, in the case of obtaining authorization of the target object, the information processing method of the present disclosure can be used to determine the behavior recognition result of the target object, and generate the behavior prompt information about the target object according to the behavior recognition result. For example, the target object can include an elderly person in a nursing home. The behavior prompt information about the behavior of the elderly person, such as the behavior prompt information about the behavior of the elderly person, can be generated according to the behavior recognition result of the elderly person. In this way, the target object in need of assistance can be found in time according to the behavior prompt information, and rescue can be taken to the target object in time to ensure the life safety of the target object.

[0062] According to an embodiment of the present disclosure, by generating the location-sensitive score map of the target object only according to the target region, the identification efficiency can be improved by avoiding identification of the full region in the target video frame. Moreover, by analyzing the behavior of the target object according to the object part feature in the location-sensitive score map, the identification granularity is improved by avoiding analysis of the overall feature of the target object, thereby improving the accuracy of identifying the behavior of the target object. Based on this, by improving the identification efficiency and the identification accuracy, the efficiency and the accuracy of the generated behavior prompt information about the target object can be improved.

[0063] Figure 3 A schematic diagram of a prompting method according to an embodiment of the present disclosure is shown schematically.

[0064] As shown in Figure 3 The prompting method of this embodiment includes operations S310-S350.

[0065] In operation S310, a target region is determined from a target video frame.

[0066] In operation S320, a location-sensitive score map about a target object is generated according to the target region.

[0067] In operation S330, an object part feature in the location-sensitive score map is analyzed to obtain a behavior recognition result about the target object.

[0068] In operation S340, in the case where the behavior recognition result is a second result, it is determined that the target object is normal.

[0069] In operation S350, in a case where the behavior recognition result is the first result, prompt information about the target object is generated to prompt the abnormal behavior of the target object.

[0070] According to an embodiment of the present disclosure, determining the target region from the target video frame comprises: performing feature extraction on the target video frame to obtain a target feature map. The target feature map is labeled according to target bounding box size information to obtain the target region, wherein the target bounding box size information is determined according to the bounding box size information, and the bounding box size information includes first bounding box size information and second bounding box size information. The first bounding box size information is set by a user, and the second bounding box size information is set by a target model.

[0071] According to an embodiment of the present disclosure, the first bounding box size information can be set by the user according to the requirements. For example, the user can set the first bounding box size information according to the size information of the target object in the historical video frame.

[0072] The second bounding box size information can include three default sizes and three aspect ratios. The three default sizes include a width of 128 pixels x height of 128 pixels, a width of 256 pixels x height of 256 pixels, and a width of 512 pixels x height of 512 pixels. The three aspect ratios include a width of 1: height of 1, a width of 1: height of 2, and a width of 2: height of 1. In this way, the above-mentioned three default sizes and three aspect ratios can be combined to obtain nine second bounding box size information. For example, 1:1 and a width of 128 pixels x height of 128 pixels are combined, and the second bounding box size information can be a width of 128 pixels x height of 128 pixels; 1:2 and a width of 128 pixels x height of 128 pixels are combined, and the second bounding box size information can be a width of 128 pixels x height of 256 pixels; 2:1 and a width of 128 pixels x height of 128 pixels are combined, and the second bounding box size information can be a width of 256 pixels x height of 128 pixels.

[0073] Due to the small difference in size of the target object, there is a problem of low labeling accuracy when only the second bounding box size information set by the target model is used to label the target feature map. Therefore, the first bounding box size information set by the user and the second bounding box size information set by the target model can be used together to label the target feature map to obtain the target region, so as to expand the size of the bounding box and improve the labeling accuracy. The target model in this example is the same as the target model obtained by training the RPN.

[0074] According to an embodiment of the present disclosure, a predetermined number of bounding boxes can be used for labeling for each target object, and the predetermined number can be 6, 7, or 8, etc. In this way, redundant bounding boxes can be avoided.

[0075] Figure 4 A schematic diagram of a target region is shown.

[0076] As shown in Figure 4 , the target feature map 400 can include a target region 410 and a background region 420. Wherein, the shaded part can represent the target region 410, and the non-shaded part can represent the background region 420. Each target region 410 can be labeled by L labeled boxes, L is a positive integer. Wherein, the background region 420 can represent the features in the target image frame other than the features of the target region. The position-sensitive score map can be obtained by processing the target feature map 400 using the classification layer and the regression layer.

[0077] According to an embodiment of the present disclosure, the target video frame can be feature extracted by using a feature extraction model to obtain a target feature map. The feature extraction model can be trained based on a residual network.

[0078] Figure 5 A schematic diagram of obtaining a target feature map is shown.

[0079] As shown in Figure 5 , in the case of complex structure of the feature extraction model, the training effect of the feature extraction model is poor. Based on this, the function to be learned can be converted from H(x) to F(x)+x by using a residual network including an identity mapping. In this way, the problem can be divided into a residual problem of multiple scales, thereby improving the training effect of the feature extraction model. And by adding a shortcut connection, the data flow can be added in a nonlinear way through the cross-layer, which does not add additional parameters and computational complexity to the residual network, and can improve the training speed and training effect of the residual network. Therefore, even if the number of network layers increases, it can also have good training effect. Based on this, the feature extraction model trained based on the residual network including the first weight layer 520 and the second weight layer 530 can be used to process the target video frame 510 to obtain the target feature map 540. Wherein, identity can represent the above-mentioned shortcut connection, and RELU (Rectified Linear Units) can represent the activation function in the feature extraction model. It should be noted that the number of first weight layer 520 and second weight layer 530 is only for illustration, and the number of first weight layer 520 and second weight layer 530 can be adjusted according to requirements.

[0080] According to an embodiment of the present disclosure, the target feature map can be labeled by using a target model according to the first label box size information and the second label box size information to obtain a target region.

[0081] According to an embodiment of the present disclosure, by using the labeling box size information set by the user and the labeling box set by the target model together for labeling, the available labeling size is expanded, so that the target feature map can be labeled using an accurate labeling box, and an accurate target region can be obtained.

[0082] According to an embodiment of the present disclosure, both the feature extraction model and the target model can label the target feature map to obtain a labeling box. For the labeling boxes labeled by the feature extraction model and the target model and located at the same position, a non-maximum suppression method can be used for processing to retain only one labeling box.

[0083] According to an embodiment of the present disclosure, the target labeling box size information is determined according to the labeling box size information by the following operations: labeling the target feature map according to the labeling box size information of the tthiteration to obtain a labeling region of the tthiteration, where t is a positive integer less than T, and T is the iteration round. From the labeling region of the tthiteration, K first center regions and J first non-center regions are determined, where K and J are positive integers. Distance information between a kthcenter region in the K first center regions and a jthnon-center region in the J first non-center regions is determined, where k is a positive integer less than K, and j is a positive integer less than J. From the J first non-center regions, a target non-center region is determined that satisfies first predetermined conditions with the kthcenter region. The labeling box size information of the t+1thiteration is determined according to the target non-center region. Difference information between the labeling box size information of the tthiteration and the labeling box size information of the t+1thiteration is determined. In the case where the difference information satisfies second predetermined conditions, the target feature map is labeled according to the labeling box size information of the t+1thiteration to obtain a target region.

[0084] According to an embodiment of the present disclosure, the labeling region of the tthiteration can include features of the target object. The labeling region of the tthiteration can be K+J. The first center region can be randomly determined from the labeling region of the tthiteration. The first non-center region can be the labeling region of the tthiteration except the first center region.

[0085] According to an embodiment of the present disclosure, the distance information can include a distance value between the kthcenter region and the jthnon-center region. The first predetermined conditions can include a first predetermined threshold.

[0086] In the case where the distance value between the kthcenter region and the jthnon-center region is less than or equal to the first predetermined threshold, the jthnon-center region can be determined as the target non-center region corresponding to the kthcenter region; in the case where the distance value between the kthcenter region and the jthnon-center region is greater than the predetermined threshold, the jthnon-center region can be discarded. The first predetermined threshold can be set according to requirements, which is not limited in the present disclosure.

[0087] According to an embodiment of the present disclosure, the difference information can include a difference value between the annotation box size information of the t th round and the annotation box size information of the t+1 th round. The difference value can include a height difference value and a width difference value. The height difference value represents a height difference between the annotation box size information of the t th round and the annotation box size information of the t+1 th round, and the width difference value represents a width difference between the annotation box size information of the t th round and the annotation box size information of the t+1 th round.

[0088] In a case where the height difference value is less than or equal to a predetermined height threshold value and the width difference value is less than or equal to a predetermined width threshold value, it can be determined that the difference information satisfies a second predetermined condition. In a case where one of the height difference value is greater than the predetermined height threshold value and the width difference value is greater than the predetermined width threshold value, it can be determined that the difference information does not satisfy the second predetermined condition. In a case where the difference information satisfies the second predetermined condition, the iteration process can be stopped. In a case where the difference information does not satisfy the second predetermined condition, t can be incremented until the difference information satisfies the second predetermined condition. The predetermined height threshold value and the predetermined width threshold value can be set according to requirements, and the present disclosure does not limit this.

[0089] According to an embodiment of the present disclosure, the annotation box size information of the 1 st round can include first annotation box size information and second annotation box size information. Thus, the target feature map can be annotated according to the annotation box size information of the 1 st round to obtain the annotation region of the 1 st round. Further, the annotation box size information of the next round can be generated in sequence.

[0090] According to an embodiment of the present disclosure, the distance information between the k th central region and the j th non-central region can be determined according to a difference value between the midpoint of the k th central region and the midpoint of the j th non-central region.

[0091] According to an embodiment of the present disclosure, the non-central region with distance information less than or equal to a first predetermined threshold value from the k th central region can be determined as a target non-central region. Thus, S target non-central regions corresponding to the k th central region can be obtained, where S is a positive integer.

[0092] According to an embodiment of the present disclosure, in a case where the target non-central region corresponding to the k th central region is one, the annotation box size information of the t+1 th round can be determined according to the size information of the target non-central region. In a case where the target non-central region corresponding to the k th central region is M, the annotation box size information of the t+1 th round can be determined according to a median value of an average of the size information of the M target non-central regions, where M is a positive integer greater than 1.

[0093] According to an embodiment of the present disclosure, the height difference value can be determined according to a height value in the annotation box size information of the t th round and a height value in the annotation box size information of the t+1 th round; and the width difference value can be determined according to a width value in the annotation box size information of the t th round and a width value in the annotation box size information of the t+1 th round.

[0094] According to an embodiment of the present disclosure, in a case where the difference information satisfies a second predetermined condition, the annotation box size information of the t+1 th round can be determined as target annotation box size information, and then the target feature map can be labeled using the target annotation box size information to obtain a target region.

[0095] According to an embodiment of the present disclosure, by determining the annotation box size information of the next round through round-by-round iteration, the size information of the annotation box can be adjusted, so that the accuracy of the annotation box can be improved, and then accurate target annotation box size information can be obtained. Based on this, the accuracy of the determined target region can be improved.

[0096] According to an embodiment of the present disclosure, the distance information between the k th central region in the K central regions and the j th non-central region in the J non-central regions is determined by: determining an intersection area between the k th central region and the j th non-central region; determining a union area between the k th central region and the j th non-central region; and determining the distance information between the k th central region and the j th non-central region according to the intersection area and the union area.

[0097] According to an embodiment of the present disclosure, the intersection area can represent an area corresponding to an overlapping part between the k th central region and the j th non-central region. The union area can represent a sum of areas between the k th central region and the j th non-central region.

[0098] According to an embodiment of the present disclosure, the calculation formula of the distance information is shown in the following formula (1) and formula (2).

[0099]

[0100]

[0101] wherein, may represent distance information between the k th central region and the j th non-central region, may represent an intersection union between the k th central region and the j th non-central region, may represent an intersection area between the k th central region and the j th non-central region, may represent a union area between the k th central region and the j th non-central region, may represent an area of the k th central region, and B jAn area of the jth non-central region can be represented.

[0102] According to an embodiment of the present disclosure, by means of the intersection area and the union area between the kth central region and the jth non-central region, accurate distance information between the kth central region and the jth non-central region can be determined.

[0103] According to an embodiment of the present disclosure, the target non-central regions are M, M is a positive integer greater than 1, and the size information of the target non-central regions includes height values and width values. According to the target non-central regions, the size information of the central region of the t+1th round is determined, including: determining the height value of the central region of the t+1th round according to the median of the height values of the M target non-central regions. The width value of the central region of the t+1th round is determined according to the median of the width values of the M target non-central regions. The size information of the central region of the t+1th round is determined according to the height value of the central region of the t+1th round and the width value of the central region of the t+1th round.

[0104] According to an embodiment of the present disclosure, the height value of the central region of the t+1th round can be determined according to the median of the average of the height values of the M target non-central regions.

[0105] According to an embodiment of the present disclosure, the width value of the central region of the t+1th round can be determined according to the median of the average of the width values of the M target non-central regions.

[0106] According to an embodiment of the present disclosure, the M target non-central regions can include a target non-central region 1 and a target non-central region 2. The height value of the target non-central region 1 can be 12mm, the width value of the target non-central region 1 can be 6mm, the height value of the target non-central region 2 can be 14mm, and the height value of the target non-central region 2 can be 4mm. Therefore, the height value of the central region of the t+1th round can be (6+7) / 2=7.5mm, and the width value of the central region of the t+1th round can be (3+2) / 2=2.5mm.

[0107] According to an embodiment of the present disclosure, by means of the median of the height value and the width value respectively, the bounding box of the next round can be continuously reduced, thereby the accuracy of the bounding box can be improved, and further the accuracy of the labeled target region can be improved.

[0108] According to an embodiment of the present disclosure, the generating, according to the target region, a position-sensitive score map about the target object comprises: generating, according to a predetermined parameter and size information of the target region, sub-region size information. Dividing the target region according to the sub-region size information to obtain Q target sub-regions, wherein Q is a positive integer greater than 1. Determining confidence information corresponding to the Q target sub-regions. Generating the position-sensitive score map according to the confidence information and the target region.

[0109] According to an embodiment of the present disclosure, the predetermined parameter can be used to determine the size ratio between the target region and the target sub-region. The predetermined parameter can include 3, 4, or 5, etc.

[0110] Figure 6 A schematic diagram of a target sub-region according to an embodiment of the present disclosure is schematically shown.

[0111] As shown in Figure 6 , in the case where the predetermined parameter is E, the height value of the target sub-region can be determined according to 1 / E of the height value in the size information of the target region, and the width value of the target sub-region can be determined according to 1 / E of the width value in the size information of the target region. Thus, E 2 target sub-regions can be obtained, wherein E 2 = Q.

[0112] According to an embodiment of the present disclosure, the confidence of each of the Q target sub-regions can be calculated by using a softmax activation function.

[0113] According to an embodiment of the present disclosure, the position-sensitive score map can be generated according to the confidence of each target sub-region and the Q target sub-regions.

[0114] According to an embodiment of the present disclosure, by dividing the target region into sub-regions and then determining the confidence according to the sub-regions, the accuracy of the determined confidence is improved, and thus the accuracy of the position-sensitive score map is improved.

[0115] According to an embodiment of the present disclosure, the determining confidence information corresponding to the Q target sub-regions comprises: pooling the Q target sub-regions to obtain a pooling region. Processing the pooling region to obtain the confidence information.

[0116] According to an embodiment of the present disclosure, the confidence information can include confidence corresponding to each of the Q target sub-regions.

[0117] According to an embodiment of the present disclosure, the trained convolutional neural network can be used to perform average pooling on each of the Q target sub-regions to obtain the pooling region.

[0118] According to an embodiment of the present disclosure, the pooling region can be processed by using a Softmax activation function to obtain confidence information.

[0119] According to an embodiment of the present disclosure, the point r c (i,j|θ) is shown in formula (3).

[0120]

[0121] Wherein, Θ can represent the network parameters of the trained convolutional neural network. x0 and y0 can represent the coordinates of the top left corner of the target region. bin(i,j) can represent the (i,j)th target sub-region in the target region. i can represent the row position of the target sub-region in the target region. j can represent the column position of the target sub-region in the target region. Z i，j，c (x,y) can represent the coordinates of the point located in the target sub-region. n can represent the number of points located in the target sub-region.

[0122] According to an embodiment of the present disclosure, by pooling the target sub-region, the required processing features are reduced, thereby improving the efficiency.

[0123] Figure 7 A schematic diagram of generating a position-sensitive score map according to an embodiment of the present disclosure is shown schematically.

[0124] As Figure 7 shown, the target region 720 can be obtained by labeling the target feature map 710. The target region 720 can also be obtained by performing convolution processing on the target feature map 710, and then labeling the convolutional target feature map 710. By convolution, part of the background features in the target feature map 710 can be removed. Further, the target region 720 can be pooled to obtain a pooling region 730. Based on this, the pooling region 730 can be processed by using a softmax activation function to obtain confidence information 740. Based on this, the position-sensitive score map 750 can be obtained according to the confidence information 740 and the target region 720. Wherein, the target region can be ROI (Region of interest, region of interest). The number of object part features in the target region 720 can be C. Based on this, since the target region 720 also includes background features, the target region 720 includes C+1 types of features. Further, the output channel corresponding to the position-sensitive score map can be E 2 (C+1). Wherein, E 2 may represent the number of target sub-regions.

[0125] For a target region with size information w x h, the size information of a target sub-region can be In the case that E is 3, nine target sub-regions can be obtained. The nine target sub-regions can be numbered to represent the spatial position information of the target object in each target sub-region, i.e., the spatial position information of the target sub-region at the top left, the target sub-region at the top middle, the target sub-region at the top right, …, and the target sub-region at the bottom left.

[0126] Based on the nine target sub-regions, a position-sensitive score map corresponding to the nine target sub-regions can be obtained.

[0127] Figure 8 A schematic diagram for obtaining confidence information according to an embodiment of the present disclosure is schematically shown.

[0128] As shown in Figure 8 , feature extraction can be performed on the target video frame 810 to obtain a target feature map 820. The target feature map 820 can be labeled to obtain a target region 830. The target region can be divided to obtain Q target sub-regions 840. Pooling can be performed on the Q target sub-regions 840 to obtain a pooling region 850. Confidence information 860 can be calculated based on the pooling region.

[0129] According to an embodiment of the present disclosure, the target video frame is F, and F is a positive integer. The object part feature is analyzed to obtain a behavior recognition result about the target object, including: determining pose information of the target object according to the object part feature corresponding to the F target video frames. The pose information is recognized to obtain the behavior recognition result about the target object.

[0130] According to an embodiment of the present disclosure, the object part feature can include a key feature point. The key feature point can include a head feature, a shoulder feature, an elbow feature, or a knee feature of the object, etc.

[0131] According to an embodiment of the present disclosure, the key feature point can be processed by using a motion recognition algorithm to obtain the pose information of the target object.

[0132] According to an embodiment of the present disclosure, the pose information can be compared with G predetermined pose information, thereby achieving recognition of the pose information, and further obtaining the behavior recognition result.

[0133] The pose information can include a pose identifier, and the predetermined pose information can include a predetermined pose identifier. The comparison of the pose information with the G predetermined pose information can include: determining that the pose identifier is compared with the G predetermined pose identifiers, and G is a positive integer.

[0134] The G pieces of predetermined posture information can include risk posture information. In a case where it is determined that the posture information is consistent with the risk posture information, a first result about the target object can be generated; and in a case where it is determined that the posture information is inconsistent with the risk posture information, a second result about the target object can be generated.

[0135] According to an embodiment of the present disclosure, by means of the object part features in the F target video frames, accurate posture information of the target object can be determined, and thus the behavior of the target object is identified by means of the posture information, improving the identification accuracy.

[0136] Based on the above information processing method, the present disclosure further provides an information processing device. The following will be described in combination with Figure 9 The device will be described in detail.

[0137] Figure 9 A structural block diagram of an information processing device according to an embodiment of the present disclosure is schematically shown.

[0138] As Figure 9 shown, the information processing device 900 of the embodiment includes a determination module 910, a first generation module 920, an analysis module 930, and a second generation module 940.

[0139] The determination module 910 is configured to determine a target region from a target video frame, wherein the target region includes a target object. In an embodiment, the determination module 910 can be configured to perform the operation S210 described above, and details are not repeated here.

[0140] The first generation module 920 is configured to generate a location-sensitive score map about the target object according to the target region, the location-sensitive score map including object part features of the target object. In an embodiment, the first generation module 920 can be configured to perform the operation S220 described above, and details are not repeated here.

[0141] The analysis module 930 is configured to analyze the object part features to obtain a behavior identification result about the target object. In an embodiment, the analysis module 930 can be configured to perform the operation S230 described above, and details are not repeated here.

[0142] The second generation module 940 is configured to generate behavior prompt information about the target object according to the behavior identification result. In an embodiment, the second generation module 940 can be configured to perform the operation S240 described above, and details are not repeated here.

[0143] According to an embodiment of the present disclosure, the determining module 910 comprises an extracting sub-module and a labeling sub-module. The extracting sub-module is configured to perform feature extraction on the target video frame to obtain a target feature map. The labeling sub-module is configured to label the target feature map according to target labeling box size information to obtain a target region, wherein the target labeling box size information is determined according to labeling box size information, and the labeling box size information comprises first labeling box size information and second labeling box size information. The first labeling box size information is set by a user, and the second labeling box size information is set by a target model.

[0144] According to an embodiment of the present disclosure, the labeling sub-module comprises a first labeling unit, a first determining unit, a second determining unit, a third determining unit, a fourth determining unit, a fifth determining unit, and a second labeling unit. The first labeling unit is configured to label the target feature map according to labeling box size information of the t-th iteration to obtain a labeled region of the t-th iteration, wherein t is a positive integer less than T, and T is the number of iterations. The first determining unit is configured to determine K first center regions and J first non-center regions from the labeled region of the t-th iteration, wherein K and J are positive integers. The second determining unit is configured to determine distance information between a k-th center region of the K first center regions and a j-th non-center region of the J first non-center regions, wherein k is a positive integer less than K, and j is a positive integer less than J. The third determining unit is configured to determine a target non-center region from the J first non-center regions, wherein the distance information between the target non-center region and the k-th center region satisfies a first predetermined condition. The fourth determining unit is configured to determine labeling box size information of the t+1-th iteration according to the target non-center region. The fifth determining unit is configured to determine difference information between the labeling box size information of the t-th iteration and the labeling box size information of the t+1-th iteration. The second labeling unit is configured to determine the labeling box size information of the t+1-th iteration as the target labeling box size information when the difference information satisfies a second predetermined condition.

[0145] According to an embodiment of the present disclosure, the second determining unit comprises a first determining sub-unit, a second determining sub-unit, and a third determining sub-unit. The first determining sub-unit is configured to determine an intersection area between the k-th center region and the j-th non-center region. The second determining sub-unit is configured to determine a union area between the k-th center region and the j-th non-center region. The third determining sub-unit is configured to determine distance information between the k-th center region and the j-th non-center region according to the intersection area and the union area.

[0146] According to an embodiment of the present disclosure, the fourth determining unit comprises a fourth determining subunit, a fifth determining subunit and a sixth determining subunit. The fourth determining subunit is configured to determine the height value of the center region in the t+1th round according to the median of the height values of the M target non-center regions; the fifth determining subunit is configured to determine the width value of the center region in the t+1th round according to the median of the width values of the M target non-center regions; and the sixth determining subunit is configured to determine the size information of the center region in the t+1th round according to the height value of the center region in the t+1th round and the width value of the center region in the t+1th round.

[0147] According to an embodiment of the present disclosure, the first generating module 920 comprises a first generating sub-module, a dividing sub-module, a first determining sub-module and a second generating sub-module. The first generating sub-module is configured to generate sub-region size information according to predetermined parameters and the size information of the target region; the dividing sub-module is configured to divide the target region according to the sub-region size information to obtain Q target sub-regions, wherein Q is a positive integer greater than 1; the first determining sub-module is configured to determine confidence information corresponding to the Q target sub-regions; and the second generating sub-module is configured to generate a location sensitive score map according to the confidence information and the target region.

[0148] According to an embodiment of the present disclosure, the first determining sub-module comprises a pooling unit and a processing unit. The pooling unit is configured to pool the Q target sub-regions to obtain a pooled region; and the processing unit is configured to process the pooled region to obtain the confidence information.

[0149] According to an embodiment of the present disclosure, the analysis module comprises a second determining sub-module and an identifying sub-module. The second determining sub-module is configured to determine the pose information of the target object according to the object part features corresponding to the F target video frames; and the identifying sub-module is configured to identify the pose information to obtain a behavior recognition result about the target object.

[0150] According to an embodiment of the present disclosure, any of the modules of the determining module 910, the first generating module 920, the analyzing module 930 and the second generating module 940 can be combined in one module, or any of the modules can be split into multiple modules. Alternatively, at least part of the functions of one or more of the modules can be combined with at least part of the functions of the other modules, and implemented in one module. According to an embodiment of the present disclosure, at least one of the determining module 910, the first generating module 920, the analyzing module 930 and the second generating module 940 can be at least partially implemented as a hardware circuit, such as a field programmable gate array (FPGA), a programmable logic array (PLA), a system on chip, a system on board, a system on package, an application specific integrated circuit (ASIC), or any other reasonable manner of integrating or packaging a circuit, etc. or implemented by hardware or firmware, or implemented in any one of software, hardware and firmware or in a proper combination of any of them. Alternatively, at least one of the determining module 910, the first generating module 920, the analyzing module 930 and the second generating module 940 can be at least partially implemented as a computer program module which can perform the corresponding functions when the computer program module is run.

[0151] Figure 10 A block diagram of an electronic device suitable for implementing the information processing method according to an embodiment of the present disclosure is schematically shown.

[0152] As shown in Figure 10 The electronic device 1000 according to an embodiment of the present disclosure includes a processor 1001 which can perform various appropriate actions and processes according to programs stored in a read only memory (ROM) 1002 or loaded from a storage portion 1008 into a random access memory (RAM) 1003. The processor 1001 can include, for example, a general purpose microprocessor (such as a CPU), an instruction set processor and / or a related chipset and / or a special purpose microprocessor (such as an application specific integrated circuit (ASIC)), etc. The processor 1001 can also include an on-board memory for cache use. The processor 1001 can include a single processing unit or multiple processing units for performing different actions of the method processes according to embodiments of the present disclosure.

[0153] In the RAM 1003, various programs and data required for the operation of the electronic device 1000 are stored. The processor 1001, the ROM 1002, and the RAM 1003 are connected to each other via the bus 1004. The processor 1001 performs various operations of the method flow according to the embodiments of the present disclosure by executing the programs in the ROM 1002 and / or the RAM 1003. It should be noted that the programs can also be stored in one or more memories other than the ROM 1002 and the RAM 1003. The processor 1001 can also perform various operations of the method flow according to the embodiments of the present disclosure by executing the programs stored in the one or more memories.

[0154] According to an embodiment of the present disclosure, the electronic device 1000 can further include an input / output (I / O) interface 1005 that is also connected to the bus 1004. The electronic device 1000 can further include one or more of the following components connected to the input / output (I / O) interface 1005: an input part 1006 including a keyboard, a mouse, etc.; an output part 1007 including a display such as a cathode ray tube (CRT), a liquid crystal display (LCD), etc., and a speaker, etc.; a storage part 1008 including a hard disk, etc.; and a communication part 1009 including a network interface card such as a LAN card, a modem, etc. The communication part 1009 performs communication processing via a network such as the Internet. A drive 1010 is also connected to the input / output (I / O) interface 1005 as necessary. A removable medium 1011 such as a magnetic disk, an optical disk, a magneto-optical disk, a semiconductor memory, etc. is attached to the drive 1010 as necessary, so that a computer program read out therefrom is installed in the storage part 1008 as necessary.

[0155] The present disclosure also provides a computer readable storage medium, which can be included in the device / apparatus / system described in the above embodiments; or can exist separately without being assembled into the device / apparatus / system. The above computer readable storage medium carries one or more programs, when the one or more programs are executed, the method according to the embodiments of the present disclosure is implemented.

[0156] According to an embodiment of the present disclosure, the computer readable storage medium can be a nonvolatile computer readable storage medium, for example, can include, but is not limited to, a portable computer diskette, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or Flash memory), a portable compact disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any appropriate combination thereof. In the present disclosure, the computer readable storage medium can be any tangible medium that contains or stores a program that can be used by or in connection with an instruction execution system, apparatus, or device. For example, according to an embodiment of the present disclosure, the computer readable storage medium can include one or more memories such as the ROM 1002 and / or the RAM 1003 described above and / or one or more memory chips other than the ROM 1002 and the RAM 1003.

[0157] Embodiments of the present disclosure also include a computer program product that includes a computer program containing program codes for executing the methods shown in the flowcharts. When the computer program product is run in a computer system, the program codes are used to cause the computer system to implement the information processing method provided by the embodiments of the present disclosure.

[0158] The above-described functions defined in the system / apparatus / module / unit of the embodiments of the present disclosure are performed when the computer program is executed by the processor 1001. According to an embodiment of the present disclosure, the system, apparatus, module, unit, etc. described above can be implemented by computer program modules.

[0159] In one embodiment, the computer program can rely on a tangible storage medium such as an optical storage device, a magnetic storage device, etc. In another embodiment, the computer program can also be transmitted, distributed, and downloaded in the form of a signal over a network medium and installed and executed by the communication part 1009 and / or installed from the detachable medium 1011. The program codes contained in the computer program can be transmitted by any appropriate network medium, including but not limited to wireless, wired, etc., or any appropriate combination thereof.

[0160] In such an embodiment, the computer program can be downloaded and installed from a network by the communication part 1009 and / or installed from the detachable medium 1011. When the computer program is executed by the processor 1001, the above-described functions defined in the system of the embodiments of the present disclosure are performed. According to an embodiment of the present disclosure, the system, device, apparatus, module, unit, etc. described above can be implemented by computer program modules.

[0161] According to embodiments of the present disclosure, program code of the computer programs provided by the embodiments of the present disclosure can be written in any combination of one or more programming languages, and specifically, these computer programs can be implemented using a high-level procedural and / or object-oriented programming language, and / or an assembly / machine language. The programming language includes, but is not limited to, a programming language such as Java, C++, Python, "C" language, or a similar programming language. The program code can be executed entirely on a user computing device, partially on a user device, partially on a remote computing device, or entirely on a remote computing device or server. In the case involving a remote computing device, the remote computing device can be connected to the user computing device through any kind of network, including a local area network (LAN) or a wide area network (WAN), or can be connected to an external computing device (for example, connected through the Internet by using an Internet service provider).

[0162] The flowcharts and block diagrams in the drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of the present disclosure. In this regard, each block in the flowcharts or block diagrams can represent a module, a segment, or a portion of code, which comprises one or more executable instructions for implementing the specified logical functions. It should also be noted that in some alternative implementations, the functions noted in the blocks can occur out of the order noted in the figures. For example, two blocks noted in succession can in fact be executed substantially concurrently or in the reverse order, depending on the functionality involved. It should also be noted that each block in the flowcharts or block diagrams, and combinations of blocks in the flowcharts or block diagrams, can be implemented by special-purpose hardware-based systems that perform the specified functions or operations, or can be implemented by a combination of special-purpose hardware and computer instructions.

[0163] It should be noted that the operations shown in the flowcharts of the embodiments of the present disclosure can be executed in any order unless otherwise specified, or unless the order of execution of the different operations is technically necessary, or unless the order of execution of the different operations is technically necessary. The execution order of multiple operations can not be distinguished, or multiple operations can be executed simultaneously.

[0164] Those skilled in the art can understand that the features described in various embodiments of the present disclosure and / or claims can be combined or / and integrated, even if such combinations or integrations are not explicitly described in the present disclosure. In particular, the features described in various embodiments of the present disclosure and / or claims can be combined and / or integrated in various combinations, without departing from the spirit and teachings of the present disclosure. All such combinations and / or integrations fall within the scope of the present disclosure.

[0165] The above describes embodiments of the present disclosure. However, these embodiments are merely for illustrative purposes, and are not intended to limit the scope of the present disclosure. Although each embodiment is described above separately, this does not mean that the measures in each embodiment cannot be used advantageously in combination. The scope of the present disclosure is defined by the appended claims and their equivalents. Those skilled in the art can make various substitutions and modifications without departing from the scope of the present disclosure, and these substitutions and modifications should all fall within the scope of the present disclosure.

Claims

1. An information processing method, comprising: determining a target region from a target video frame, wherein the target region comprises a target object; generating a location-sensitive score map about the target object according to the target region, wherein the location-sensitive score map comprises object part features of the target object; analyzing the object part features to obtain a behavior recognition result about the target object; generating behavior prompt information about the target object according to the behavior recognition result; wherein the determining the target region from the target video frame comprises: performing feature extraction on the target video frame to obtain a target feature map; labeling the target feature map according to target bounding box size information to obtain the target region, wherein the target bounding box size information is determined according to bounding box size information, and the bounding box size information comprises first bounding box size information and second bounding box size information, the first bounding box size information is set by a user, and the second bounding box size information is set by a target model; wherein the target bounding box size information is determined according to the bounding box size information by the following operations: labeling the target feature map according to the t-th round of bounding box size information to obtain the t-th round of labeled region, wherein t is a positive integer less than T, and T is an iteration round; determining K first center regions and J first non-center regions from the t-th round of labeled region, wherein K and J are positive integers; determining distance information between a k-th center region in the K first center regions and a j-th non-center region in the J first non-center regions, wherein k is a positive integer less than K, and j is a positive integer less than J; determining a target non-center region from the J first non-center regions, wherein distance information between the target non-center region and the k-th center region satisfies a first predetermined condition; determining the t+1-th round of bounding box size information according to the target non-center region; determining difference information between the t-th round of bounding box size information and the t+1-th round of bounding box size information; in a case where the difference information satisfies a second predetermined condition, determining the t+1-th round of bounding box size information as the target bounding box size information.

2. The method of claim 1, wherein, The determining distance information between the k-th center region in the K first center regions and the j-th non-center region in the J first non-center regions comprises: determining an intersection area between the k-th center region and the j-th non-center region; determining a union area between the k-th center region and the j-th non-center region; determining distance information between the k-th center region and the j-th non-center region according to the intersection area and the union area.

3. The method of claim 1, wherein, The target non-center region is M, M is a positive integer greater than 1, and size information of the target non-center region comprises a height value and a width value; The determining size information of the t+1-th round of center region according to the target non-center region comprises: determine the height value of the center region in the t+1th round according to the median of the height values of the M target non-center regions; determine the width value of the center region in the t+1th round according to the median of the width values of the M target non-center regions; determine the size information of the center region in the t+1th round according to the height value of the center region in the t+1th round and the width value of the center region in the t+1th round.

4. The method of claim 1, wherein, The generating of the position-sensitive score map of the target object according to the target region comprises: generating sub-region size information according to predetermined parameters and size information of the target region; dividing the target region according to the sub-region size information to obtain Q target sub-regions, wherein Q is a positive integer greater than 1; determining confidence information corresponding to the Q target sub-regions; generating the position-sensitive score map according to the confidence information and the target region.

5. The method of claim 4, wherein, The determination of the confidence information corresponding to the Q target sub-regions comprises: pooling the Q target sub-regions to obtain a pooling region; processing the pooling region to obtain the confidence information.

6. The method of claim 1, wherein, The target video frame is F, and F is a positive integer; The analysis of the object part feature to obtain the behavior recognition result of the target object comprises: determining pose information of the target object according to the object part feature corresponding to the F target video frames; recognizing the pose information to obtain the behavior recognition result of the target object.

7. An information processing apparatus, comprising: a determination module configured to determine a target region from a target video frame, wherein the target region comprises a target object; a first generation module configured to generate a position-sensitive score map of the target object according to the target region, wherein the position-sensitive score map comprises an object part feature of the target object; an analysis module configured to analyze the object part feature to obtain a behavior recognition result of the target object; a second generation module configured to generate behavior prompt information of the target object according to the behavior recognition result; wherein the determination of the target region from the target video frame comprises: performing feature extraction on the target video frame to obtain a target feature map; and performing labeling on the target feature map according to target bounding box size information to obtain the target region, wherein the target bounding box size information is determined according to bounding box size information, the bounding box size information comprises first bounding box size information and second bounding box size information, the first bounding box size information is set by a user, and the second bounding box size information is set by a target model. The target annotation box size information is determined according to the annotation box size information by the following operations: according to the annotation box size information of the tthiteration, the target feature map is annotated to obtain a labeled region of the tthiteration, where t is a positive integer less than T, and T is the iteration round; K first center regions and J first non-center regions are determined from the labeled region of the tthiteration, where K and J are positive integers; distance information between a kthcenter region in the K first center regions and a jthnon-center region in the J first non-center regions is determined, where k is a positive integer less than K, and j is a positive integer less than J; a target non-center region that satisfies a first predetermined condition in the distance information between the kthcenter region and the J first non-center regions is determined from the J first non-center regions; annotation box size information of a (t+1) thiteration is determined according to the target non-center region; difference information between the annotation box size information of the tthiteration and the annotation box size information of the (t+1) thiteration is determined; and in a case where the difference information satisfies a second predetermined condition, the annotation box size information of the (t+1) thiteration is determined as the target annotation box size information. 8.An electronic device, comprising: one or more processors; a storage device for storing one or more programs, wherein the one or more programs, when executed by the one or more processors, cause the one or more processors to perform the method according to any one of claims 1-6. 9.A computer-readable storage medium having stored thereon executable instructions that, when executed by a processor, cause the processor to perform the method according to any one of claims 1-6. 10.A computer program product comprising a computer program that, when executed by a processor, implements the method according to any one of claims 1-6.

Images