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

Abstract

The application relates to a target object identification method and device, computer equipment and a storage medium, relates to the field of image recognition, and can be applied to the field of financial technology or other related fields.The method comprises the following steps: extracting point cloud data of a candidate object from data collected for a target region; determining action sequence information of the candidate object according to the point cloud data of the candidate object, wherein the action sequence information comprises an action sequence constructed by at least one action; determining a target value corresponding to the candidate object according to the action sequence information of the candidate object, wherein the target value is used for representing the probability that the candidate object is a target object; and determining a target object identification result for the candidate object according to the target value corresponding to the candidate object. The method can improve the identification efficiency of the target object.

Description

Technical Field

[0001] This application relates to the field of image recognition, and in particular to a method, apparatus, computer device, and storage medium for identifying target objects. Background Technology

[0002] Branches serve as the primary channel for commercial banks to serve their customers and are a window into the bank's external service image. In the daily operations of commercial banks, there is often a need to screen personnel entering the bank.

[0003] In existing technologies, personnel identification is usually carried out by manually reviewing video recordings of service outlets, but this method is inefficient. Summary of the Invention

[0004] Therefore, it is necessary to provide a target object identification method, apparatus, computer device, and storage medium to address the aforementioned technical problems.

[0005] Firstly, this application provides a method for identifying a target object. The method includes:

[0006] Extract point cloud data of candidate objects from data collected from the target area;

[0007] Based on the point cloud data of the candidate object, the action sequence information of the candidate object is determined, and the action sequence information includes an action sequence constructed from at least one action;

[0008] Based on the action sequence information of the candidate object, a target value corresponding to the candidate object is determined, and the target value is used to characterize the probability that the candidate object is the target object;

[0009] Based on the target value corresponding to the candidate object, the target object recognition result for the candidate object is determined.

[0010] In one embodiment, determining the target value corresponding to the candidate object based on the action sequence information of the candidate object includes:

[0011] The target action is determined from the action sequence information;

[0012] The target action is matched with preset action information to obtain the matching result corresponding to the target action;

[0013] If the matching result indicates that there is a preset action in the preset action information that matches the target action, then the target action is taken as the starting action;

[0014] Based on the starting action and the number of consecutive preset actions following the starting action in the action sequence information, construct the action group of the candidate object;

[0015] Based on the action group of the candidate object, determine the target value corresponding to the candidate object.

[0016] In one embodiment, the method further includes:

[0017] If the matching result indicates that there is no preset action in the preset action information that matches the target action, the action that is adjacent to the target action and located after the target action in the action sequence information is taken as the target action, and the step of matching the target action with the preset action information to obtain the matching result corresponding to the target action is repeated.

[0018] In one embodiment, determining the target value corresponding to the candidate object based on the action group of the candidate object includes:

[0019] Obtain the historical action group corresponding to the historical target object, wherein the historical target object is the target object determined during the historical target object identification process;

[0020] For any of the historical target objects, a first feature similarity between the candidate object and the historical target object is determined based on the historical action group corresponding to the historical target object and the action group of the candidate object.

[0021] The target value corresponding to the candidate object is determined based on the first feature similarity between the candidate object and each of the historical target objects.

[0022] In one embodiment, determining the target object recognition result for the candidate object based on the target value corresponding to the candidate object includes:

[0023] If the target value corresponding to the candidate object is greater than the target value threshold, the motion trajectory of the candidate object is determined based on the point cloud data of the candidate object.

[0024] Based on the motion trajectory of the candidate object, determine the first trajectory target value;

[0025] If the first trajectory target value is greater than the trajectory target value threshold, the candidate object is taken as the target object.

[0026] In one embodiment, determining the first trajectory target value based on the motion trajectory of the candidate object includes:

[0027] Based on the movement trajectory of the candidate object, determine the dwell time of the candidate object at each preset location, and based on the dwell time of the candidate object at each preset location, determine the second trajectory target value;

[0028] Based on the motion trajectory of the candidate object, the distance between the candidate object and each object to be compared is determined, and a third trajectory target value is determined based on the distance between the candidate object and each object to be compared. The objects to be compared are all the candidate objects except the candidate object.

[0029] The first trajectory target value is determined based on the second trajectory target value and the third trajectory target value.

[0030] In one embodiment, determining the second trajectory target value based on the dwell time of the candidate object at each preset location includes:

[0031] For any of the preset locations, if the candidate's stay time at the preset location is less than a time threshold, the target location value corresponding to the candidate at the preset location is determined based on the stay time; or,

[0032] If the dwell time is greater than or equal to the duration threshold, the location target value of the candidate object at the preset location is determined based on the action sequence information of the candidate object at the preset location.

[0033] The second trajectory target value of the candidate object is determined based on the location target value corresponding to each of the preset locations.

[0034] In one embodiment, determining the third trajectory target value based on the distance between the candidate object and each object to be compared includes:

[0035] For any of the objects to be compared, based on the motion trajectory of the candidate object, determine the duration for which the distance between the candidate object and the object to be compared is less than a distance threshold, and based on the duration for which the distance between the candidate object and the object to be compared is less than the distance threshold, determine the target value of the candidate object relative to the object to be compared.

[0036] Based on the target values ​​of the candidate objects for each of the objects to be compared, determine the fourth trajectory target value corresponding to the candidate object;

[0037] Based on the motion trajectory of the candidate object, determine the number of times the distance between the candidate object and each of the objects to be compared is less than the distance threshold, and determine the fifth trajectory target value based on the number of times the distance between the candidate object and each of the objects to be compared is less than the distance threshold;

[0038] The third trajectory target value is determined based on the fourth trajectory target value and the fifth trajectory target value.

[0039] Secondly, this application also provides a target object identification device. The device includes:

[0040] The extraction module is used to extract point cloud data of candidate objects from data collected for the target area;

[0041] The first determining module is used to determine the action sequence information of the candidate object based on the point cloud data of the candidate object, wherein the action sequence information includes an action sequence constructed from at least one action;

[0042] The second determining module is used to determine the target value corresponding to the candidate object based on the action sequence information of the candidate object, wherein the target value is used to characterize the probability that the candidate object is the target object;

[0043] The third determining module is used to determine the target object recognition result for the candidate object based on the target value corresponding to the candidate object.

[0044] In one embodiment, the second determining module is further configured to:

[0045] The target action is determined from the action sequence information;

[0046] The target action is matched with preset action information to obtain the matching result corresponding to the target action;

[0047] If the matching result indicates that there is a preset action in the preset action information that matches the target action, then the target action is taken as the starting action;

[0048] Based on the starting action and the number of consecutive preset actions following the starting action in the action sequence information, construct the action group of the candidate object;

[0049] Based on the action group of the candidate object, determine the target value corresponding to the candidate object.

[0050] In one embodiment, the device further includes:

[0051] The processing module is configured to, when the matching result indicates that there is no preset action in the preset action information that matches the target action, take the action that is adjacent to the target action and located after the target action in the action sequence information as the target action, and repeat the step of matching the target action with the preset action information to obtain the matching result corresponding to the target action.

[0052] In one embodiment, the second determining module is further configured to:

[0053] Obtain the historical action group corresponding to the historical target object, wherein the historical target object is the target object determined during the historical target object identification process;

[0054] For any of the historical target objects, a first feature similarity between the candidate object and the historical target object is determined based on the historical action group corresponding to the historical target object and the action group of the candidate object.

[0055] The target value corresponding to the candidate object is determined based on the first feature similarity between the candidate object and each of the historical target objects.

[0056] In one embodiment, the third determining module is further configured to:

[0057] If the target value corresponding to the candidate object is greater than the target value threshold, the motion trajectory of the candidate object is determined based on the point cloud data of the candidate object.

[0058] Based on the motion trajectory of the candidate object, determine the first trajectory target value;

[0059] If the first trajectory target value is greater than the trajectory target value threshold, the candidate object is taken as the target object.

[0060] In one embodiment, the third determining module is further configured to:

[0061] Based on the movement trajectory of the candidate object, determine the dwell time of the candidate object at each preset location, and based on the dwell time of the candidate object at each preset location, determine the second trajectory target value;

[0062] Based on the motion trajectory of the candidate object, the distance between the candidate object and each object to be compared is determined, and a third trajectory target value is determined based on the distance between the candidate object and each object to be compared. The objects to be compared are all the candidate objects except the candidate object.

[0063] The first trajectory target value is determined based on the second trajectory target value and the third trajectory target value.

[0064] In one embodiment, the third determining module is further configured to:

[0065] For any of the preset locations, if the candidate's stay time at the preset location is less than a time threshold, the target location value corresponding to the candidate at the preset location is determined based on the stay time; or,

[0066] If the dwell time is greater than or equal to the duration threshold, the location target value of the candidate object at the preset location is determined based on the action sequence information of the candidate object at the preset location.

[0067] The second trajectory target value of the candidate object is determined based on the location target value corresponding to each of the preset locations.

[0068] In one embodiment, the third determining module is further configured to:

[0069] For any of the objects to be compared, based on the motion trajectory of the candidate object, determine the duration for which the distance between the candidate object and the object to be compared is less than a distance threshold, and based on the duration for which the distance between the candidate object and the object to be compared is less than the distance threshold, determine the target value of the candidate object relative to the object to be compared.

[0070] Based on the target values ​​of the candidate objects for each of the objects to be compared, determine the fourth trajectory target value corresponding to the candidate object;

[0071] Based on the motion trajectory of the candidate object, determine the number of times the distance between the candidate object and each of the objects to be compared is less than the distance threshold, and determine the fifth trajectory target value based on the number of times the distance between the candidate object and each of the objects to be compared is less than the distance threshold;

[0072] The third trajectory target value is determined based on the fourth trajectory target value and the fifth trajectory target value.

[0073] Thirdly, this application also provides a computer device. The computer device includes a memory and a processor, the memory storing a computer program, and the processor executing the computer program to implement any of the methods described above.

[0074] Fourthly, this application also provides a computer-readable storage medium. The computer-readable storage medium stores a computer program thereon, which, when executed by a processor, implements any of the above methods.

[0075] Fifthly, this application also provides a computer program product. The computer program product includes a computer program, which, when executed by a processor, implements any of the methods described above.

[0076] The aforementioned target object recognition method, apparatus, computer equipment, and storage medium, by extracting point cloud data of candidate objects and constructing action sequence information of candidate objects, can determine the target value, i.e., the probability that a candidate object is the target object, and then determine whether the candidate object is the target object based on the target value. The embodiments of this application construct action sequence information of candidate objects from point cloud data, and then automatically determine whether a candidate object is the target object based on the action sequence of the candidate object, which can improve the efficiency and accuracy of target object recognition. Attached Figure Description

[0077] Figure 1 This is a flowchart illustrating a target object identification method in one embodiment;

[0078] Figure 2 This is a flowchart illustrating step 106 in one embodiment;

[0079] Figure 3 This is a flowchart illustrating step 210 in one embodiment;

[0080] Figure 4 This is a flowchart illustrating step 108 in one embodiment;

[0081] Figure 5 This is a flowchart illustrating step 406 in one embodiment;

[0082] Figure 6 This is a flowchart illustrating step 502 in one embodiment;

[0083] Figure 7 This is a flowchart illustrating step 504 in one embodiment;

[0084] Figure 8 This is a schematic diagram of a target object recognition system in one embodiment;

[0085] Figure 9 This is a schematic diagram illustrating the construction of a historical target object database in one embodiment;

[0086] Figure 10 This is a schematic diagram of a target object identification method in one embodiment;

[0087] Figure 11 This is a structural block diagram of a target object recognition device in one embodiment;

[0088] Figure 12 This is an internal structural diagram of a computer device in one embodiment. Detailed Implementation

[0089] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0090] In one embodiment, such as Figure 1 As shown, a target object recognition method is provided. This embodiment illustrates the application of this method to a terminal. It is understood that this method can also be applied to a server, and further to a system including both a terminal and a server, and is implemented through interaction between the terminal and the server. In this embodiment, the method includes the following steps:

[0091] Step 102: Extract point cloud data of candidate objects from the data collected for the target area.

[0092] In this embodiment, the target area can be a pre-defined area, such as a specific location or the area that an image acquisition device can capture. When applied to target personnel identification in a bank branch, the target area can be the customer-facing area of ​​the bank branch. The data collected for the target area can be video or image data captured by an image acquisition device, or radar point cloud data collected by radar, etc. This embodiment does not specifically limit this. Candidate objects can be any object to be identified entering the target area, such as any person entering the target area. By clustering the point cloud data, each object within the target area can be determined, and then point cloud data representing candidate objects can be filtered from each object. The point cloud data of candidate objects can include the point cloud data of the candidate object itself and the point cloud data within a certain range around the candidate object. For example, if the candidate object is a person, the point cloud data of the candidate object can include the point cloud data of that person, as well as the point cloud data of other people within a certain range around the candidate object, and the point cloud data of items near the candidate object.

[0093] It should be noted that if the data collected for the target area is video or image data, then an image acquisition device capable of obtaining the distance between the object and the camera, such as a depth camera, should be used to capture the aforementioned video or image data so as to extract the point cloud data of the candidate objects based on the video or image data. If the data collected for the target area is radar point cloud data, then the point cloud data of the candidate objects can be directly extracted from the radar point cloud data. This application does not specifically limit the method for extracting the point cloud data of candidate objects; any method that can extract the point cloud data of candidate objects based on the data collected for the target area is applicable to this application.

[0094] Meanwhile, if the data collected for the target area is video data or image data, staff can be excluded from the candidates by means of facial recognition, clothing recognition, etc., before extracting the point cloud data of the candidates, so as to reduce the number of candidates that need to be compared later.

[0095] Step 104: Determine the action sequence information of the candidate object based on the point cloud data of the candidate object. The action sequence information includes an action sequence constructed from at least one action.

[0096] In this embodiment, the action sequence information may include all actions of the candidate object from the moment it enters the target area to the current moment. The actions of the candidate object can be constructed in real time based on continuously extracted point cloud data. This embodiment does not specifically limit the method for constructing the actions of the candidate object; any method that can construct candidate objects based on point cloud data is applicable to this embodiment. For example, a neural network specific to the application scenario of this embodiment can be pre-trained, and the actions of the candidate object can be obtained by inputting point cloud data into the neural network during the application.

[0097] Step 106: Determine the target value corresponding to the candidate object based on the action sequence information of the candidate object. The target value is used to characterize the probability that the candidate object is the target object.

[0098] In this embodiment of the application, the target value is used to characterize the probability that the candidate object is the target object. The larger the target value, the higher the probability that the candidate object is the target object.

[0099] For example, the target value corresponding to the candidate object can be determined based on each action in the action sequence information. For instance, if a preset number of consecutive actions in the action sequence information match actions that have a high probability of belonging to the target object, a score can be assigned to the candidate object based on the degree of matching, and this score can be used as the target value corresponding to the candidate object. This application does not specifically limit this aspect.

[0100] Step 108: Determine the target object recognition result for the candidate object based on the target value corresponding to the candidate object.

[0101] In this embodiment, a target value threshold can be preset. When the target value corresponding to a candidate object is greater than or equal to the target value threshold, the candidate object can be determined as the target object; when the target value corresponding to a candidate object is less than the target value threshold, the candidate object can be determined as not being the target object. The specific value of the target value threshold can be determined by those skilled in the art according to actual needs. For example, when the accuracy requirement for the target object recognition result is high, a higher target value threshold can be set; when the accuracy requirement for the target object recognition result is not high, a lower target value threshold can be set.

[0102] Once a candidate has been identified as the target, staff can be prompted to verify the target by sending notification messages.

[0103] The target object recognition method provided in this application extracts point cloud data of candidate objects and constructs action sequence information of the candidate objects to determine a target value, that is, the probability that a candidate object is a target object. Then, it determines whether a candidate object is a target object based on the target value. This application's method constructs action sequence information of candidate objects from point cloud data and automatically determines whether a candidate object is a target object based on the action sequence, thereby improving the efficiency and accuracy of target object recognition.

[0104] In one embodiment, such as Figure 2 As shown, in step 106, the target value corresponding to the candidate object is determined based on the action sequence information of the candidate object, including:

[0105] Step 202: Determine the target action from the action sequence information.

[0106] Step 204: Match the target action with the preset action information to obtain the matching result corresponding to the target action.

[0107] Step 206: If there is a preset action that matches the target action in the preset action information represented by the matching result, the target action is taken as the starting action.

[0108] In this embodiment, the target action is an action within the action sequence information. The preset action information consists of actions that are pre-defined and have a high probability of belonging to the target object. The specific preset action information can be set by those skilled in the art based on the different target objects. For example, if the target object is the object that executes the target event, the target object will typically execute actions 1 and 2 when executing the target event; therefore, the preset action information may include actions 1 and 2, etc.

[0109] For example, the first action in the action sequence information can be directly used as the target action, or the action corresponding to the action sequence information at the time point that is more likely to appear to match the preset action information can be used as the target action, etc. The embodiments of this application do not specifically limit this.

[0110] After determining the target action, it can be matched against each preset action one by one. For example, the feature similarity between the target action and each preset action can be determined individually. When the feature similarity between the target action and a preset action is greater than or equal to a feature similarity threshold, it can be determined that the target action matches the preset action, and the target action can be used as the starting action. If the feature similarity between the target action and the preset action is less than the feature similarity threshold, it can be determined that the target action does not match the preset action. In this case, the matching process can continue to match the target action with the next preset action.

[0111] Step 208: Based on the starting action and the number of consecutive preset actions following the starting action in the action sequence information, construct the action group of the candidate object.

[0112] Step 210: Determine the target value corresponding to the candidate object based on the action group of the candidate object.

[0113] In this embodiment of the application, after determining the starting action, a set of action groups consisting of a preset number of consecutive actions following the starting action and the starting action itself can be constructed as candidate action groups. The preset number can be a pre-defined number, and its specific value can be determined by those skilled in the art according to their needs. For example, if the starting action is to walk towards other customers, and the consecutive actions following the starting action are to talk to other customers, show paper to other customers, stand up, and walk towards other customers, then when the preset number is 3, the action group includes 4 actions: walking towards other customers, talking to other customers, showing paper to other customers, and standing up; when the preset number is 4, the action group includes 5 actions: walking towards other customers, talking to other customers, showing paper to other customers, standing up, and walking towards other customers.

[0114] After obtaining the action group of the candidate object, the target value corresponding to the candidate object can be obtained by matching the action group with the action group that has a high probability of belonging to the target object.

[0115] The target object recognition method provided in this application can match target actions in action sequence information with preset actions. When a target action matches any preset action, it constructs a group of action candidates based on the target action and the actions following it, and determines the target value of the candidate object based on the group of action. This application can determine the probability that a candidate object is a target object based on a group of action actions, i.e., several consecutive actions of the candidate object. Therefore, the judgment criteria are more reasonable, and the accuracy of target object identification can be improved.

[0116] In one embodiment, the method further includes:

[0117] If there is no preset action that matches the target action in the preset action information represented by the matching result, the action that is adjacent to the target action and located after the target action in the action sequence information is taken as the target action, and the steps of matching the target action with the preset action information are repeated to obtain the matching result corresponding to the target action are repeated.

[0118] In this embodiment, if the target action does not match any of the preset actions, the next action after the target action (the first action after the target action) can be used as the target action, and the action can continue to be matched with each preset action. If the first action after the target action still does not match any of the preset actions, the second action after the target action can be used as the target action, and the action can be matched with each preset action... and so on, until an action is successfully matched with a preset action, or there are no other actions in the action sequence information.

[0119] For example, if the action sequence of a candidate object is: walking (first action), interacting with the ticket dispenser (second action), walking (third action), sitting down (fourth action), then the first action can be used as the target action and matched with each preset action. When the first action fails to match (i.e., it does not match any preset action), the second action can be used as the target action and matched with each preset action. And so on. When the fourth action also fails to match, since there is no fifth action in the action sequence information, the matching can be stopped, and the candidate object can be temporarily marked as a non-target object until a new action of the candidate object can be constructed based on the point cloud data.

[0120] The target object identification method provided in this application can, even when the target action does not match any of the preset actions, continue to use the next action adjacent to the target action as the target action and repeat the step of matching the target action with the preset action information. This application embodiment can determine the probability of a candidate object being the target object based on action groups, that is, several consecutive actions of the candidate object. Therefore, the judgment criteria are more reasonable and can improve the accuracy of target object identification.

[0121] In one embodiment, such as Figure 3 As shown, in step 210, the target value corresponding to the candidate object is determined based on the candidate object's action group, including:

[0122] Step 302: Obtain the historical action group corresponding to the historical target object. The historical target object is the target object identified during the historical target object identification process.

[0123] Step 304: For any historical target object, determine the first feature similarity between the action group of the candidate object and the historical action group based on the historical action group corresponding to the historical target object and the action group of the candidate object.

[0124] Step 306: Determine the target value corresponding to the candidate object based on the similarity of the first feature between the candidate object and each historical target object.

[0125] In this embodiment, after identifying historical target objects during the historical target object identification process, the historical action groups corresponding to the historical target objects can be stored in a database. After determining the action groups of candidate objects, the first feature similarity between the action groups of the candidate objects and the historical action groups corresponding to each historical target object can be determined, and the target value corresponding to the candidate object can be determined based on each first feature similarity. This embodiment does not specifically limit the method of determining the first feature similarity between the action groups of candidate objects and historical action groups. For example, the feature similarity between the actions in the action groups of candidate objects and the corresponding actions in the historical action groups can be determined (for example, the similarity between the first action in the action groups of candidate objects and the first action in the historical action groups, the similarity between the second action in the action groups of candidate objects and the second action in the historical action groups, etc. can be determined), and the first feature similarity between the action groups of candidate objects and historical action groups can be obtained by summing, weighted summing, or averaging the feature similarities.

[0126] Furthermore, a feature similarity threshold can be preset. After obtaining multiple first feature similarities corresponding to the action group of the candidate object, the target value corresponding to the candidate object can be obtained by comparing each first feature similarity with the similarity threshold. For example, the average value of each first feature similarity can be taken, and the target value corresponding to the candidate object can be determined based on the difference between the average value and the feature similarity threshold. Alternatively, the target value corresponding to the candidate object can be determined based on the proportion of first feature similarities with values ​​greater than or equal to the similarity threshold in all first feature similarities. This application embodiment does not specifically limit this.

[0127] Once a candidate object is identified as the target object, its action group can be added to the database as a historical action group in the subsequent target object identification process.

[0128] The target object identification method provided in this application can determine the target value corresponding to a candidate object based on the similarity between the action group of a candidate object and the historical action group corresponding to a historical target object. Since the historical action group is the action group corresponding to the target object determined during the historical target object identification process, if the action group of a candidate object is similar to the historical action group, the candidate object has a high probability of being the target object. Using the similarity between the action group of a candidate object and the historical action group as the criterion for judging whether a candidate object is a target object can improve the accuracy of target object determination.

[0129] In one embodiment, such as Figure 4 As shown, in step 108, the target object recognition result for the candidate object is determined based on the target value corresponding to the candidate object, including:

[0130] Step 402: If the target value corresponding to the candidate object is greater than the target value threshold, determine the motion trajectory of the candidate object based on the point cloud data of the candidate object.

[0131] Step 404: Determine the first trajectory target value based on the motion trajectory of the candidate object.

[0132] Step 406: If the first trajectory target value is greater than the trajectory target value threshold, the candidate object is taken as the target object.

[0133] In this embodiment of the application, a target value threshold can be preset. If the target value corresponding to the candidate object is greater than the target value threshold, it means that the candidate object is likely to be the target object. At this time, the movement trajectory of the candidate object can be used to verify whether the candidate object is the target object.

[0134] For example, the point cloud data of candidate objects at a certain moment can be clustered to determine the coordinates of the candidate objects in three-dimensional space at that moment, and the motion trajectory of the candidate objects can be determined based on the coordinates of the candidate objects at each moment. Depending on whether the motion trajectory of the candidate objects is normal (e.g., whether it follows the typical motion trajectory of customers, whether it intersects with the business area, etc.), a score, i.e., the first trajectory target value, can be assigned to the motion trajectory of the candidate objects. The larger the first trajectory target value, the greater the probability that the candidate object is the target object. A threshold for the first trajectory target value can also be preset; if the first trajectory target value is greater than the threshold, the candidate object can be identified as the target object.

[0135] The target object recognition method provided in this application embodiment can verify whether a candidate object is a target object by the motion trajectory of the candidate object. Only when the target value of the candidate object is greater than the target value threshold and the target value of the first trajectory is also greater than the target value of the first trajectory, the candidate object is determined to be a target object. Therefore, the probability of misjudging the target object can be reduced.

[0136] In one embodiment, such as Figure 5 As shown, in step 406, the first trajectory target value is determined based on the motion trajectory of the candidate object, including:

[0137] Step 502: Based on the movement trajectory of the candidate object, determine the dwell time of the candidate object at each preset location, and based on the dwell time of the candidate object at each preset location, determine the second trajectory target value.

[0138] In this embodiment, the preset location can be a business area within a bank branch. The duration of a candidate's stay at each preset location can be determined based on the candidate's movement trajectory and the boundary range of each preset location within the target area. For example, if a candidate's movement trajectory enters the boundary of a preset location at 10:05 and leaves the boundary at 10:17, the duration of the candidate's stay at the preset location can be determined to be 12 minutes. Conversely, if the candidate's movement trajectory never enters the boundary of a preset location, the duration of the candidate's stay at the preset location can be determined to be 0 minutes.

[0139] For example, since it can be determined that the candidate has not conducted any business when the candidate's dwell time is 0, the probability that the candidate is the target object is relatively high. Therefore, the second trajectory target value of the candidate can be set to a larger value. When the candidate's dwell time is less than the time usually required to conduct business, it can also be said that the candidate has not conducted any business. Therefore, a time threshold can be preset. When the candidate's dwell time is less than the time threshold, the second trajectory target value of the candidate can also be set to a larger value. When the candidate's dwell time is greater than or equal to the time threshold, it means that the candidate has a high probability of conducting business. Therefore, the second trajectory target value can be set to a smaller value. This application embodiment does not make specific limitations in this regard.

[0140] Step 504: Based on the motion trajectory of the candidate object, determine the distance between the candidate object and each object to be compared, and determine the target value of the third trajectory based on the distance. The objects to be compared are the objects other than the candidate objects among all the candidate objects.

[0141] Step 506: Determine the first trajectory target value based on the second trajectory target value and the third trajectory target value.

[0142] In this embodiment of the application, the distance between the candidate object and the object to be compared can be determined based on the motion trajectory of the candidate object and the motion trajectory of the object to be compared. For example, the distance between the candidate object and the object to be compared at a certain moment can be determined based on the coordinates of the candidate object at a certain moment and the coordinates of the object to be compared at that moment, and the third trajectory target value of the candidate object can be determined based on the distance.

[0143] For example, since candidate objects that repeatedly approach different objects to be compared have a high probability of being the target object, a distance threshold can be preset. When the distance between a candidate object and the object to be compared is less than the distance threshold, a score can be assigned to the candidate object. When the distance between the candidate object and another different object to be compared is less than the distance threshold again, a score can be assigned to the candidate object again, and so on, until the sum of all scores is used as the target value of the third trajectory. Alternatively, the target value of the third trajectory can also be determined based on the duration for which the distance between the candidate object and the object to be compared is less than the distance threshold, etc., and this application embodiment does not specifically limit this.

[0144] Furthermore, the first trajectory target value can be obtained based on the second and third trajectory target values. For example, the first trajectory target value can be obtained by summing the second and third trajectory target values. Alternatively, weights can be assigned to the second and third trajectory target values ​​respectively, and the first trajectory target value can be obtained by weighted summing of the second and third trajectory target values.

[0145] The target object identification method provided in this application can determine a second trajectory target value based on the duration of a candidate object's stay at a preset location, and determine a third trajectory target value based on the distance between the candidate object and each object to be compared. Then, a first trajectory target value is determined based on the second and third trajectory target values. Since the larger the second and third trajectory target values, the higher the probability that the candidate object is the target object, and the larger the first trajectory target value obtained from the second and third trajectory target values, the higher the probability that the candidate object is the target object when the first trajectory target value is greater than a first trajectory target value threshold, thus reducing the probability of misidentifying the target object.

[0146] In one embodiment, such as Figure 6 As shown, in step 502, the second trajectory target value is determined based on the duration of the candidate object's stay at each preset location, including:

[0147] Step 602: For any preset location, if the candidate's stay time at the preset location is less than a time threshold, determine the target value of the candidate at the preset location based on the stay time; or,

[0148] Step 604: If the dwell time is greater than or equal to the duration threshold, determine the location target value of the candidate object at the preset location based on the action sequence information corresponding to the action of the candidate object at the preset location.

[0149] Step 606: Determine the second trajectory target value of the candidate object based on the target value corresponding to each preset location.

[0150] In this embodiment, a time threshold can be preset, which represents the time typically required to complete a transaction. When the dwell time is less than the time threshold, a shorter dwell time indicates a higher probability that the candidate has not completed the transaction, so the target location value for the candidate at that preset location can be set relatively high. Conversely, a longer dwell time indicates that although the candidate's dwell time is less than the time threshold, there is still a possibility that the candidate may complete the transaction, so the target location value for the candidate at that preset location can be set relatively low.

[0151] If the dwell time is greater than or equal to the duration threshold, the candidate's actions at the preset location can be determined from the candidate's movement sequence information to more accurately determine whether the candidate actually conducted business at the preset location. For example, if the preset location is a bank counter, and the candidate does not sit down or talk to the bank teller at the preset location, it can be determined that there is a high probability that the candidate did not conduct business at the preset location.

[0152] Furthermore, the target location value for a candidate at a preset location can be determined based on whether the candidate has conducted business at that location. For example, since there is a higher probability that a candidate who has entered the boundary of a preset location but has not conducted business is the target, the target location value can be set higher if it can be determined from the candidate's actions that the candidate has not conducted business at the preset location; conversely, if it can be determined from the candidate's actions that the candidate has conducted business at the preset location, the target location value can be set to 0.

[0153] Furthermore, the second trajectory target value can be determined based on the location target value corresponding to the candidate object at each preset location. For example, the second trajectory target value can be obtained by summing the location target values ​​corresponding to the candidate object at each preset location. This application embodiment does not specifically limit this.

[0154] The target object recognition method provided in this application embodiment can determine whether a candidate object is conducting business at a preset location based on the candidate object's corresponding actions at the preset location, and determine a second trajectory target value based on whether the candidate object is conducting business, thus improving the accuracy of the determined second trajectory target value.

[0155] In one embodiment, such as Figure 7 As shown, in step 504, the third trajectory target value is determined based on the distance between the candidate object and each object to be compared, including:

[0156] Step 702: For any object to be compared, determine the duration during which the distance between the candidate object and the object to be compared is less than a distance threshold based on the motion trajectory of the candidate object, and determine the target value of the candidate object relative to the object to be compared based on the duration during which the distance between the candidate object and the object to be compared is less than the distance threshold.

[0157] Step 704: Determine the fourth trajectory target value corresponding to the candidate object based on the target value of the candidate object for each object to be compared.

[0158] Step 706: Based on the motion trajectory of the candidate object, determine the number of times the distance between the candidate object and each object to be compared is less than the distance threshold, and determine the fifth trajectory target value based on the number of times the distance between the candidate object and each object to be compared is less than the distance threshold.

[0159] Step 708: Determine the target value of the third trajectory based on the target values ​​of the fourth and fifth trajectories.

[0160] In this embodiment, since the candidate object is more likely to be the target object when the distance between the candidate object and the object to be compared is less than the distance threshold for a longer period of time, or when the distance between the candidate object and different objects to be compared is less than the distance threshold more often, the target object of the candidate object is more likely to be the target object. Therefore, the target value of the candidate object relative to the object to be compared can be determined based on the duration of the distance between the candidate object and a certain object to be compared being less than the distance threshold (the longer the duration of the distance between the candidate object and the object to be compared being less than the distance threshold, the larger the target value of the candidate object relative to the object to be compared should be). And the fourth trajectory target value is determined based on the target value of the candidate object relative to all objects to be compared. For example, the fourth trajectory target value can be obtained by summing the target values ​​of the candidate object relative to each object to be compared. This embodiment does not specifically limit this.

[0161] Furthermore, the target value of the fifth trajectory can be determined based on the number of times the distance between the candidate object and each object to be compared is less than the distance threshold. The more times the distance between the candidate object and each object to be compared is less than the distance threshold, the larger the target value of the fifth trajectory should be.

[0162] Furthermore, the third trajectory target value can be obtained based on the fourth and fifth trajectory target values. For example, the third trajectory target value can be obtained by summing the fourth and fifth trajectory target values. Alternatively, weights can be assigned to the fourth and fifth trajectory target values ​​respectively, and the third trajectory target value can be obtained by weighted summing of the fourth and fifth trajectory target values.

[0163] The target object identification method provided in this application can determine a fourth trajectory target value based on the duration during which the distance between a candidate object and each comparison object is less than a distance threshold, and determine a fifth trajectory target value based on the number of times the distance between the candidate object and each comparison object is less than a distance threshold. Then, a third trajectory target value is determined based on the fourth and fifth trajectory target values. Since the larger the fourth and fifth trajectory target values, the higher the probability that the candidate object is the target object, and the larger the third trajectory target value obtained from the fourth and fifth trajectory target values, the higher the probability that the candidate object is the target object when the first trajectory target value obtained from the third and second trajectory target values ​​is greater than the first trajectory target value threshold, thus reducing the probability of misidentifying the target object.

[0164] To enable those skilled in the art to better understand the embodiments of this application, the embodiments of this application are described below through specific examples.

[0165] In this embodiment of the application, the above-described target object identification method can be applied to... Figure 8 In the target object recognition system shown, the image acquisition device / radar device 801 is used to collect data for the target area. When this embodiment of the application is applied to the target personnel recognition in a bank branch, the target area is the branch lobby area. In order to ensure the richness and accuracy of the point cloud data of the candidate objects, the image acquisition device / radar device should be set up to ensure that the acquisition range of each image device / radar device can fully cover the target area.

[0166] The main control unit 802 is the control center of the target object recognition system, which controls and coordinates the normal operation of each unit.

[0167] The video stream parsing unit 803 is used to acquire video data acquired by the image acquisition device and to perform decoding operations on the video data.

[0168] The historical archive processing unit 804 is used to process historical data containing target objects and obtain the action group of the historical target objects.

[0169] The real-time detection unit 805 is used to perform real-time detection on the data collected for the target area and detect the target objects present in the target area.

[0170] The parameter setting unit 806 is used to set business parameters, such as target value threshold, duration threshold, distance threshold, etc.

[0171] The operation display unit 807 is used to provide timely warnings for target objects detected in real time and to send text messages or email notifications to staff.

[0172] Storage unit 808 is used to store data such as historical action groups, preset actions, and real-time warnings.

[0173] This application embodiment can also perform target object recognition using video data captured by an image acquisition device. For example... Figure 9 The diagram illustrates a flowchart for building a historical target object database. For example, after staff confirm the presence of a target object at a target location within a certain time period, video data for that time period can be extracted, and OpenCV (an open-source computer vision library) can be used to split the video data into image data of video frames. Further, the YOLO (You Only LookOnce, an object detection algorithm) v5 algorithm can be used to detect and segment candidate objects from all image data, obtaining sub-image data containing candidate objects. Candidate objects whose size does not meet the filtering criteria can then be filtered out from the sub-image data. For example, the filtering criteria could be that the aspect ratio of the candidate object is greater than 2, the height is greater than 128 pixels, and the width is greater than 96 pixels.

[0174] Furthermore, based on the image data corresponding to the target object, sub-image data corresponding to the target object can be filtered from each sub-image data. The similarity between the image data corresponding to the target object and each sub-image data can be determined, and sub-image data with a similarity greater than a similarity threshold can be identified as sub-image data corresponding to the target object, i.e., historical sub-image data of the target object. For example, the similarity threshold can be 0.7.

[0175] like Figure 10 As shown, in practical applications, real-time video data of the target area can be acquired, and OpenCV can be used to convert the video data into image data of video frames. Furthermore, the YOLO (You Only LookOnce, an object detection algorithm) v5 algorithm can be used to detect and segment candidate objects from all image data, obtaining sub-image data containing candidate objects, and then filtering the sub-image data. Filtering criteria can also include an aspect ratio greater than 2, a height greater than 128 pixels, and a width greater than 96 pixels for the candidate objects.

[0176] Furthermore, features can be extracted from the sub-image data and the sub-image data of historical target objects in the historical target object database, and the similarity between the two can be determined. If the similarity is greater than the similarity threshold, it indicates that the match is successful. At this time, an early warning message can be output to notify the staff to verify.

[0177] Alternatively, DeepSORT (Deep Simple Online and Realtime Tracking, a multi-target tracking algorithm) can be used to construct the trajectory of the candidate object for each image acquisition device, and the trajectory of multiple image acquisition devices can be used to re-identify the personnel, draw the real-time trajectory information of the candidate object throughout the entire process, and then determine whether the candidate object is the target object based on the trajectory information. The judgment logic is as follows: (1) The time spent in the target area exceeds a certain duration; (2) The trajectory of the candidate object (excluding internal employees and security guards by their clothing) is within a certain distance (e.g., 0.5m) and the time exceeds a certain duration (e.g., 1 minute) with the trajectory of other objects, and this situation occurs multiple times, exceeding the set number of times (e.g., 10 times); (3) The candidate object has not conducted any business, and the judgment logic is that the trajectory of the candidate object does not intersect with the business handling areas such as ATMs or counters. These rules can be flexibly combined and used together.

[0178] When a target object is detected, a warning message can be output in a timely manner, which may include image data, trajectory data and other data of the candidate object.

[0179] This application embodiment processes historical and real-time video data of the target area to obtain a historical target object database and detect target objects in real time. When a target object appears, an early warning is issued and staff are notified to verify it, which can effectively improve the efficiency of target object identification.

[0180] It should be understood that although the steps in the flowcharts of the embodiments described above are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the embodiments described above may include multiple steps or multiple stages. These steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the steps or stages of other steps.

[0181] Based on the same inventive concept, this application also provides a target object identification device for implementing the target object identification method described above. The solution provided by this device is similar to the implementation scheme described in the above method; therefore, the specific limitations in one or more target object identification device embodiments provided below can be found in the limitations of the target object identification method described above, and will not be repeated here.

[0182] In one embodiment, such as Figure 11 As shown, a target object recognition device 1100 is provided, including: an extraction module 1102, a first determination module 1104, a second determination module 1106, and a third determination module 1108, wherein:

[0183] Extraction module 1102 is used to extract point cloud data of candidate objects from data collected for the target area;

[0184] The first determining module 1104 is used to determine the action sequence information of the candidate object based on the point cloud data of the candidate object, wherein the action sequence information includes an action sequence constructed from at least one action;

[0185] The second determining module 1106 is used to determine the target value corresponding to the candidate object based on the action sequence information of the candidate object, wherein the target value is used to characterize the probability that the candidate object is the target object;

[0186] The third determining module 1108 is used to determine the target object recognition result for the candidate object based on the target value corresponding to the candidate object.

[0187] The target object recognition device provided in this application extracts point cloud data of candidate objects and constructs action sequence information of the candidate objects to determine the target value, that is, the probability that the candidate object is the target object. Then, it determines whether the candidate object is the target object based on the target value. This application embodiment constructs action sequence information of candidate objects from point cloud data and automatically determines whether the candidate object is the target object based on the action sequence, which can improve the recognition efficiency and accuracy of target objects.

[0188] In one embodiment, the second determining module 1106 is further configured to:

[0189] The target action is determined from the action sequence information;

[0190] The target action is matched with preset action information to obtain the matching result corresponding to the target action;

[0191] If the matching result indicates that there is a preset action in the preset action information that matches the target action, then the target action is taken as the starting action;

[0192] Based on the starting action and the number of consecutive preset actions following the starting action in the action sequence information, construct the action group of the candidate object;

[0193] Based on the action group of the candidate object, determine the target value corresponding to the candidate object.

[0194] In one embodiment, the device further includes:

[0195] The processing module is configured to, when the matching result indicates that there is no preset action in the preset action information that matches the target action, take the action that is adjacent to the target action and located after the target action in the action sequence information as the target action, and repeat the step of matching the target action with the preset action information to obtain the matching result corresponding to the target action.

[0196] In one embodiment, the second determining module 1106 is further configured to:

[0197] Obtain the historical action group corresponding to the historical target object, wherein the historical target object is the target object determined during the historical target object identification process;

[0198] For any of the historical target objects, a first feature similarity between the candidate object and the historical target object is determined based on the historical action group corresponding to the historical target object and the action group of the candidate object.

[0199] The target value corresponding to the candidate object is determined based on the first feature similarity between the candidate object and each of the historical target objects.

[0200] In one embodiment, the third determining module 1108 is further configured to:

[0201] If the target value corresponding to the candidate object is greater than the target value threshold, the motion trajectory of the candidate object is determined based on the point cloud data of the candidate object.

[0202] Based on the motion trajectory of the candidate object, determine the first trajectory target value;

[0203] If the first trajectory target value is greater than the trajectory target value threshold, the candidate object is taken as the target object.

[0204] In one embodiment, the third determining module 1108 is further configured to:

[0205] Based on the movement trajectory of the candidate object, determine the dwell time of the candidate object at each preset location, and based on the dwell time of the candidate object at each preset location, determine the second trajectory target value;

[0206] Based on the motion trajectory of the candidate object, the distance between the candidate object and each object to be compared is determined, and a third trajectory target value is determined based on the distance between the candidate object and each object to be compared. The objects to be compared are all the candidate objects except the candidate object.

[0207] The first trajectory target value is determined based on the second trajectory target value and the third trajectory target value.

[0208] In one embodiment, the third determining module 1108 is further configured to:

[0209] For any of the preset locations, if the candidate's stay time at the preset location is less than a time threshold, the target location value corresponding to the candidate at the preset location is determined based on the stay time; or,

[0210] If the dwell time is greater than or equal to the duration threshold, the location target value of the candidate object at the preset location is determined based on the action sequence information of the candidate object at the preset location.

[0211] The second trajectory target value of the candidate object is determined based on the location target value corresponding to each of the preset locations.

[0212] In one embodiment, the third determining module 1108 is further configured to:

[0213] For any of the objects to be compared, based on the motion trajectory of the candidate object, determine the duration for which the distance between the candidate object and the object to be compared is less than a distance threshold, and based on the duration for which the distance between the candidate object and the object to be compared is less than the distance threshold, determine the target value of the candidate object relative to the object to be compared.

[0214] Based on the target values ​​of the candidate objects for each of the objects to be compared, determine the fourth trajectory target value corresponding to the candidate object;

[0215] Based on the motion trajectory of the candidate object, determine the number of times the distance between the candidate object and each of the objects to be compared is less than the distance threshold, and determine the fifth trajectory target value based on the number of times the distance between the candidate object and each of the objects to be compared is less than the distance threshold;

[0216] The third trajectory target value is determined based on the fourth trajectory target value and the fifth trajectory target value.

[0217] Each module in the aforementioned target object recognition device can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device, or stored in the memory of a computer device as software, so that the processor can call and execute the operations corresponding to each module.

[0218] In one embodiment, a computer device is provided, which may be a server, and its internal structure diagram may be as follows: Figure 12 As shown, the computer device includes a processor, memory, and a network interface connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The network interface is used to communicate with external terminals via a network connection. When executed by the processor, the computer program implements a target object identification method.

[0219] Those skilled in the art will understand that Figure 12 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0220] In one embodiment, a computer device is provided, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps in the above-described method embodiments.

[0221] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon that, when executed by a processor, implements the steps in the above method embodiments.

[0222] In one embodiment, a computer program product is provided, including a computer program that, when executed by a processor, implements the steps in the above method embodiments.

[0223] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, data stored, data displayed, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties.

[0224] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the embodiments described above. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM). The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, etc., and are not limited to these.

[0225] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0226] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of this patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this application should be determined by the appended claims.

Claims

1. A method for identifying a target object, characterized in that, The method includes: Extract point cloud data of candidate objects from data collected from the target area; Based on the point cloud data of the candidate object, the action sequence information of the candidate object is determined, and the action sequence information includes an action sequence constructed from at least one action; Based on the action sequence information of the candidate object, a target value corresponding to the candidate object is determined, and the target value is used to characterize the probability that the candidate object is the target object; Based on the target value corresponding to the candidate object, determine the target object recognition result for the candidate object; The step of determining the target value corresponding to the candidate object based on the action sequence information of the candidate object includes: The target action is determined from the action sequence information; The target action is matched with preset action information to obtain the matching result corresponding to the target action; If the matching result indicates that there is a preset action in the preset action information that matches the target action, then the target action is taken as the starting action; Based on the starting action and the number of consecutive preset actions following the starting action in the action sequence information, construct the action group of the candidate object; Based on the action group of the candidate object, determine the target value corresponding to the candidate object.

2. The method according to claim 1, characterized in that, The method further includes: If the matching result indicates that there is no preset action in the preset action information that matches the target action, the action that is adjacent to the target action and located after the target action in the action sequence information is taken as the target action, and the matching of the target action with the preset action information is repeated to obtain the matching result corresponding to the target action.

3. The method according to claim 1 or 2, characterized in that, The step of determining the target value corresponding to the candidate object based on the action group of the candidate object includes: Obtain the historical action group corresponding to the historical target object, wherein the historical target object is the target object determined during the historical target object identification process; For any of the historical target objects, a first feature similarity between the candidate object and the historical target object is determined based on the historical action group corresponding to the historical target object and the action group of the candidate object. The target value corresponding to the candidate object is determined based on the first feature similarity between the candidate object and each of the historical target objects.

4. The method according to claim 1, characterized in that, The step of determining the target object recognition result for the candidate object based on the target value corresponding to the candidate object includes: If the target value corresponding to the candidate object is greater than the target value threshold, the motion trajectory of the candidate object is determined based on the point cloud data of the candidate object. Based on the motion trajectory of the candidate object, determine the first trajectory target value; If the first trajectory target value is greater than the trajectory target value threshold, the candidate object is taken as the target object.

5. The method according to claim 4, characterized in that, The step of determining the first trajectory target value based on the motion trajectory of the candidate object includes: Based on the movement trajectory of the candidate object, determine the dwell time of the candidate object at each preset location, and based on the dwell time of the candidate object at each preset location, determine the second trajectory target value; Based on the motion trajectory of the candidate object, the distance between the candidate object and each object to be compared is determined, and a third trajectory target value is determined based on the distance between the candidate object and each object to be compared. The objects to be compared are all the candidate objects except the candidate object. The first trajectory target value is determined based on the second trajectory target value and the third trajectory target value.

6. The method according to claim 5, characterized in that, The step of determining the second trajectory target value based on the dwell time of the candidate object at each preset location includes: For any of the preset locations, if the candidate's stay time at the preset location is less than a time threshold, the target location value corresponding to the candidate at the preset location is determined based on the stay time; or, If the dwell time is greater than or equal to the duration threshold, the location target value of the candidate object at the preset location is determined based on the action sequence information of the candidate object at the preset location. The second trajectory target value of the candidate object is determined based on the location target value corresponding to each of the preset locations.

7. The method according to claim 5, characterized in that, The step of determining the third trajectory target value based on the distance between the candidate object and each object to be compared includes: For any of the objects to be compared, based on the motion trajectory of the candidate object, determine the duration for which the distance between the candidate object and the object to be compared is less than a distance threshold, and based on the duration for which the distance between the candidate object and the object to be compared is less than the distance threshold, determine the target value of the candidate object relative to the object to be compared. Based on the target values ​​of the candidate objects for each of the objects to be compared, determine the fourth trajectory target value corresponding to the candidate object; Based on the motion trajectory of the candidate object, determine the number of times the distance between the candidate object and each of the objects to be compared is less than the distance threshold, and determine the fifth trajectory target value based on the number of times the distance between the candidate object and each of the objects to be compared is less than the distance threshold; The third trajectory target value is determined based on the fourth trajectory target value and the fifth trajectory target value.

8. A target object recognition device, characterized in that, The device includes: The extraction module is used to extract point cloud data of candidate objects from data collected for the target area; The first determining module is used to determine the action sequence information of the candidate object based on the point cloud data of the candidate object, wherein the action sequence information includes an action sequence constructed from at least one action; The second determining module is used to determine the target value corresponding to the candidate object based on the action sequence information of the candidate object, wherein the target value is used to characterize the probability that the candidate object is the target object; The third determining module is used to determine the target object recognition result for the candidate object based on the target value corresponding to the candidate object; The second determining module is specifically used to: determine the target action from the action sequence information; The target action is matched with preset action information to obtain the matching result corresponding to the target action; If the matching result indicates that there is a preset action in the preset action information that matches the target action, then the target action is taken as the starting action; Based on the starting action and the number of consecutive preset actions following the starting action in the action sequence information, construct the action group of the candidate object; Based on the action group of the candidate object, determine the target value corresponding to the candidate object.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 7.

11. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 7.

Images