A single-vision-based drop point automatic reporting method and system

Abstract

The application discloses a kind of based on monocular vision's drop point automatic reporting method and system, this method includes: based on target camera real-time acquisition being thrown object movement scene image;Based on the image frame after first time analysis target being thrown object movement, the first time is the time when target being thrown object starts to appear in real-time acquisition movement scene image frame;In second time trigger processor starts the storage process of acquisition image frame, the second time is the time when the movement position condition of target being thrown object meets preset storage condition;Based on the image frame analysis target being thrown object drop point position.The application in based on monocular vision system acquisition image carries out the drop point analysis of being thrown object, avoids artificial search drop point, and inefficiently obtains drop point position information by using ruler measurement.Meanwhile, monocular vision system is used, avoids the high cost of using binocular vision system.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of machine vision, in particular to a landing point automatic reporting method and system based on monocular vision. BACKGROUND

[0002] The discus, javelin, iron discus and other field events in sports events need to report the athlete's performance in real time. The traditional method is to draw a scale on the target area, and then find the landing point by manual measurement after the object is thrown down, so as to obtain the landing point position information. This method is low in efficiency and cannot adapt to the rhythm of rapid training and rapid competition. Some large-scale events have installed binocular vision measurement systems, but the cost is high and cannot be popularized in small training venues. SUMMARY

[0003] In view of the problems existing in the prior art, the present application provides a landing point automatic reporting method and system based on monocular vision, which can report the performance after the object is thrown down without manual intervention and has the advantage of low cost. The technical scheme is as follows:

[0004] In a first aspect, a landing point automatic reporting method based on monocular vision is provided, comprising the following steps:

[0005] Real-time acquisition of a target camera of a thrown object motion scene image;

[0006] Analysis of the motion of the target thrown object based on the image frame acquired after the first time, the first time being the time when the target thrown object first appears in the real-time acquired motion scene image frame;

[0007] Triggering the processor to start the storage process of the acquired image frame at the second time, the second time being the time when the motion position of the target thrown object meets the preset storage condition;

[0008] Analysis of the landing point position of the target thrown object based on the stored image frame.

[0009] In a possible implementation, before the target camera real-time acquisition of the thrown object motion scene image, the target camera is calibrated to determine the real coordinates of each pixel point in the image acquired by the target camera in the motion scene.

[0010] In a possible implementation, the first time is obtained by using frame difference method to detect whether the target thrown object appears in the motion scene image acquired at the current time based on the motion scene image acquired at the current time and the motion scene image acquired at the previous time, and if so, the current time is the first time.

[0011] In a possible implementation, the motion of the target thrown object is analyzed based on the image frame captured after the first time, including:

[0012] The frame difference method is used to analyze the selected region in the image frame based on the image frame captured after the first time, to determine whether the target thrown object appears in the selected region.

[0013] The storage process of the captured image frame is triggered by the processor at the second time, including:

[0014] When the target thrown object appears in the selected region in the image frame, the storage process of the captured image frame is triggered by the processor.

[0015] In a possible implementation, the selected region is a region composed of a plurality of continuous rows of pixels in the image frame.

[0016] In a possible implementation, the frame difference method is used to analyze the selected region in the image frame based on the image frame captured after the first time, to determine whether the target thrown object appears in the selected region, including the following steps:

[0017] The initial motion parameters of the target thrown object at the first time are analyzed, and the theoretical landing time of the target thrown object is predicted based on the initial motion parameters, and the initial motion parameters include the initial motion displacement, the initial speed, and the initial speed direction;

[0018] The position range region of the target thrown object at a time before the preset time length of the predicted value of the theoretical landing time is taken as the selected region;

[0019] The frame difference method is used to analyze the selected region in the image frame based on the image frame captured after the first time, to determine whether the target thrown object appears in the selected region.

[0020] In a possible implementation, the preset time length is determined according to the size of the resource occupied by the image storage per unit time of the target camera.

[0021] In a possible implementation, the landing position of the target thrown object is analyzed based on the stored image frame, including:

[0022] The landing image frame corresponding to the landing moment of the target thrown object is obtained based on the stored captured image frame.

[0023] The position of the target thrown object is identified and analyzed based on the landing image frame.

[0024] In a possible implementation, the landing image frame corresponding to the landing moment of the target thrown object is obtained, including:

[0025] acquire the target thrown object position on the acquisition image frame in time sequence based on the acquisition image frames stored by the present time;

[0026] generate the motion direction feature data of the target thrown object based on the target thrown object positions of the adjacent two image frames;

[0027] judge the relevance of the motion direction feature data before the present time and the motion direction feature data of the present time and the next time based on the motion direction feature data;

[0028] if the relevance is less than a preset relevance threshold, determine that the target thrown object has the landing phenomenon, and determine the image frames acquired at the present time and the next time as the landing image frames.

[0029] In a possible implementation, the identifying and analyzing the position of the target thrown object based on the landing image frames comprises:

[0030] determine that the target thrown object does not have the trajectory interruption based on the motion direction feature data generated based on the actual target thrown object positions of the next time and the present time;

[0031] acquire a first position of the target thrown object represented by the image frame of the present time and a second position represented by the image frame of the next time in the landing image frames;

[0032] determine that the first position is the landing point or the landing point is between the first position and the second position based on the height data of the first position.

[0033] In a possible implementation, the target camera is installed in a region close to the end where the target thrown object lands, and is erected close to the ground.

[0034] In a possible implementation, the target camera is erected at a height of 20 cm from the ground.

[0035] In a second aspect, a landing point automatic reporting system based on monocular vision is provided, comprising:

[0036] a motion scene image acquisition module, configured to acquire a thrown object motion scene image in real time based on a target camera;

[0037] a target motion parameter analysis module, configured to analyze the motion of the target thrown object based on the image frames acquired after a first time, the first time being the time when the target thrown object starts to appear in the motion scene image frame acquired in real time;

[0038] a motion target image storage module, configured to trigger the processor to start a storage process of the acquisition image frames at a second time, the second time being the time when the motion position of the target thrown object meets a preset storage condition;

[0039] A landing position determination module is configured to determine the landing position of the target object based on the stored image frames.

[0040] The single-vision-based landing point automatic reporting method and system has the following advantages:

[0041] 1. In the present application, the image of the moving scene is collected, and the landing point of the target object is analyzed based on the image collected by the single-vision system, thereby avoiding the low efficiency of manually searching for the landing point and measuring the landing point position information with a ruler. Meanwhile, the single-vision system is used, thereby avoiding the high cost of using a binocular vision system.

[0042] 2. In the present application, the time when the target object appears in the image frame of the moving scene is recorded as the first time, the motion parameters of the target object are analyzed based on the image of the moving scene collected at the first time, and the second time for triggering the processor to start the storage process of the collected image frame is determined, so as to reasonably utilize the limited storage space in the FPGA.

[0043] 3. In the present application, the first time is obtained by calculating whether there is a pixel region with large changes in the collected images at two time points through the frame difference method. If there is, the target object appears in the collected image of the moving scene. This method is simple and fast in calculation, reduces the time delay of image collection and real-time analysis during the high-speed motion of the target object, and further provides the training results for the training athletes such as shot put, javelin and discus.

[0044] 4. In the present application, the storage process is started only when the target object moves to a certain space region in the world coordinate system, and the selected region in the image frame of the target camera is imaged in the selected region in the image frame. In the present application, the selected region in the image frame is set as a region composed of a plurality of continuous rows of pixels. Based on this, the frame difference method is only needed for the plurality of continuous rows of pixels in the image frame in the present application, thereby further simplifying the calculation amount, improving the analysis speed of the image frame, and reducing the time delay of providing the training results for the training athletes such as shot put, javelin and discus. BRIEF DESCRIPTION OF DRAWINGS

[0045] Figure 1 is a flowchart of a single-vision-based landing point automatic reporting method in the embodiments of the present application.

[0046] Figure 2 is a structural block diagram of a single-vision-based landing point automatic reporting system in the embodiments of the present application. DETAILED DESCRIPTION

[0047] In order to make the purposes, technical solutions and advantages of the present application clearer, the present application will be further described in detail below with reference to the drawings, and the described embodiments should not be regarded as limitations on the present application. All other embodiments obtained by those of ordinary skill in the art without creative labor fall within the scope of protection of the present application.

[0048] The embodiment of the present application provides a landing point automatic reporting method based on monocular vision, comprising the following steps:

[0049] S1: collecting a target thrown object motion scene image in real time based on a target camera;

[0050] S2: analyzing the motion condition of the target thrown object based on an image frame collected after a first time, wherein the first time is the time when the target thrown object starts to appear in the real-time collected motion scene image frame;

[0051] S3: triggering the processor to start the storage process of the collected image frame at a second time, wherein the second time is the time when the motion condition of the target thrown object meets a preset condition;

[0052] S4: analyzing the landing position of the target thrown object based on the stored image frame.

[0053] In the sports events such as shot put, javelin, discus and other field sports, for the landing point measurement of the target thrown objects such as shot put, javelin and discus, in the embodiment of the present application, the landing point of the target thrown object is analyzed based on the image collected by the monocular vision system by collecting the motion scene image of the target thrown object, which avoids the low efficiency of manually searching for the landing point and measuring the landing point position information with a ruler. At the same time, the monocular vision system is used, which avoids the high cost of using the binocular vision system.

[0054] In the embodiment of the present application, only the part of the motion trajectory video image at the time when the target thrown object lands is saved. In the specific implementation process, the motion trajectory image collected in real time is stored only when the target thrown object is at the preset position of the entire motion trajectory. Of course, in order to improve the accuracy of the landing point position analysis, the processor is triggered to start the image storage process when the target thrown object is at the preset position of the entire motion trajectory, and the video image of a period of time before the storage triggering time is saved.

[0055] Further, before the above step S1, that is, before the target camera collects the motion scene image of the target thrown object in real time, the target camera is calibrated to determine the real coordinates of each pixel point in the image collected by the target camera in the motion scene.

[0056] In the embodiment of the present application, the sensor parameter of the target camera is 1920X1080 60fps, of course, the higher the frame rate of the sensor, the more accurate the analysis result. The target camera is calibrated to obtain a mapping table between the camera pixel coordinates and the real coordinates, and the static picture is obtained through WIFI transmission, the position information is marked, the calibration parameters are calculated and saved to the processor.

[0057] Further, the method for obtaining the first time in step S2 comprises: using frame difference method to detect whether the target thrown object appears in the motion scene image collected at the current time based on the motion scene image collected at the current time and the motion scene image collected at the previous time, if yes, the current time is the first time.

[0058] In the embodiment of the present application, whether there is a pixel region with large changes in the images collected at the two times is calculated by the frame difference method, if there is, the target thrown object appears in the collected motion scene image, the calculation method is simple and fast, the time delay of image collection and real-time analysis in the high-speed motion process of the target thrown object is reduced, and further, the training results of the training athletes such as shot put, javelin and discus can be provided in time.

[0059] Further, the method for analyzing the motion of the target thrown object based on the image frames collected after the first time in step S2 comprises:

[0060] S21: based on the image frames collected after the first time, frame difference method is used to analyze the selected region in the image frames, and whether the target thrown object appears in the selected region is judged;

[0061] The step S3 of triggering the processor to start the storage process of the collected image frames at the second time comprises:

[0062] S31: when the target thrown object appears in the selected region in the image frames, it is the second time, and the processor is triggered to start the storage process of the collected image frames.

[0063] Considering the limited storage space in the FPGA, only a part of the whole motion trajectory image of the target thrown object is stored, and further, in the determination of the time of starting the storage process, the present application starts to collect when the target thrown object moves to a certain space region in the world coordinate system, specifically, the space region in the world coordinate system is imaged in the selected region of the collected image frames of the target camera, in an embodiment, the selected region in the image frames is set as a region composed of a plurality of continuous rows of pixels in the image frames, on this basis, only the frame difference method is used for the plurality of continuous rows of pixels in the image frames, which further simplifies the calculation amount, improves the analysis speed of the image frames, and reduces the time delay of providing the training results of the training athletes such as shot put, javelin and discus.

[0064] Further, in step S21, based on the image frames collected after the first time, frame difference method analysis is performed on the selected region in the image frames to determine whether the selected region has the target thrown object, including the following steps:

[0065] S211: analyze the initial motion parameters of the target thrown object at the first time, and based on the initial motion parameters, predict the theoretical landing time of the target thrown object, the initial motion parameters including initial motion displacement, initial speed and initial speed direction;

[0066] S212: the position range area of the target thrown object at a time before the preset time of the predicted value of the theoretical landing time is taken as the selected region;

[0067] S213: based on the image frames collected after the first time, frame difference method analysis is performed on the selected region in the image frames to determine whether the selected region has the target thrown object.

[0068] In the embodiment of the application, the entire motion trajectory image of the thrown object in the entire motion scene is not stored, but only the part of the motion trajectory collected image containing the time before and after the landing of the thrown object and the preset time before landing is stored.

[0069] The preset time is determined according to the size of the image storage resource occupied by the target camera in unit time. Specifically, in the starting time of the storage process, the size of the image storage resource occupied by the target camera in unit time is determined according to the image acquisition frame rate, image resolution and other factors of the target camera, and the capacity of the FPGA that can be used to store the motion trajectory image data of the target thrown object is determined according to the preset capacity, and the length of time for storing the motion trajectory image of the target thrown object is determined, that is, the preset time is obtained. Further, the motion parameters of the target thrown object are analyzed according to the previously collected motion trajectory image, and the theoretical landing time of the target thrown object is predicted, and the storage process of the collected image frame is started at the time of the preset time before the predicted value of the theoretical landing time, that is, the target moves to the selected region. In the embodiment of the application, the amount of data stored by the FPGA for the motion trajectory image of the target thrown object is reduced, and the selected region is determined flexibly according to the actual motion state parameters of the target thrown object, which ensures that the selected region is best adapted to different motion processes of different target thrown objects, and improves the effectiveness and accuracy of the second time when the selected region appears the target thrown object.

[0070] Further, step S4: based on the stored image frames, the landing position of the target thrown object is analyzed, including:

[0071] S41: based on the stored collected image frames, the landing image frame corresponding to the landing time of the target thrown object is obtained;

[0072] S42: Identify the position of the target thrown object based on the landing image frame.

[0073] In step S41, the landing image frame corresponding to the time when the target thrown object lands is obtained, including:

[0074] S411: Based on the acquisition image frames stored up to the current time, the target thrown object position on the acquisition image frames is obtained in chronological order;

[0075] S412: Generate the motion direction feature data of the target thrown object based on the positions of the target thrown object in the adjacent two image frames;

[0076] S413: Determine the relevance of the motion direction feature data before the current time and the motion direction feature data of the current time and the next time based on the motion direction feature data;

[0077] S414: If the relevance is less than a preset relevance threshold, it is determined that the target thrown object has landed, and the image frames collected at the current time and the next time are determined as the landing image frames.

[0078] In the embodiments of the present application, if the motion direction between the target thrown object positions of the current time and the previous time suddenly changes compared with the motion direction of the target thrown object before the current time, it is determined that the target thrown object has landed. Specifically, the landing can be trajectory termination, or trajectory rebound, etc.

[0079] In step S412, the motion direction feature data of the target thrown object is generated based on the positions of the target thrown object in the adjacent two image frames, including:

[0080] Generate a three-element feature vector based on the horizontal coordinates and vertical coordinates of the positions of the target thrown object in the adjacent two image frames, the three-element feature vector including the horizontal coordinate change, the vertical coordinate change, and the direction of the line connecting the positions of the target thrown object in the adjacent two image frames.

[0081] In step S413, the relevance can be analyzed by the following steps:

[0082] Based on the time series of the three-element feature vectors obtained from the acquisition image frames before the current time, analyze the change rule of the three-element feature vectors over time, and predict the three-element feature vector generated from the positions of the target thrown object at the next time and the current time based on the change rule;

[0083] Determine the relevance based on the similarity between the three-element feature vector generated from the actual positions of the target thrown object at the next time and the current time and the predicted three-element feature vector.

[0084] In step S42, the position of the target object is identified based on the landing image frame obtained in step S41, including:

[0085] S420: Based on the motion direction feature data, i.e., the three-element feature vector, generated from the actual target object position at the next moment and the current moment, it is determined that the target object has not stopped the trajectory;

[0086] S421: Obtain the first position of the target object represented by the current moment image frame and the second position represented by the next moment image frame in the landing image frame;

[0087] S422: Based on the height data of the first position, it is determined that the first position is the landing position or the landing position is between the first position and the second position.

[0088] Considering that the target camera may not accurately capture the image frame at the moment when the object falls, it is possible to capture one frame of motion trajectory image before and after the actual landing, so in the embodiment of the present application, based on the determined landing image frame, the first position and the second position representing the target object are obtained by target recognition. When the height data of the first position is inconsistent with the ground height, it is determined that the landing position is between the first position and the second position, and the method for searching the landing position between the first position and the second position includes:

[0089] Based on the position of the target object in the image frame captured before the landing image frame, a first motion trajectory curve is fitted, and based on the position of the target object in the image frame captured after the landing image frame, a second motion trajectory curve is fitted. The intersection position of the first motion trajectory curve and the second motion trajectory curve is taken as the actual landing position.

[0090] In the embodiment of the present application, in order to realize continuous analysis of multiple training processes, when it is determined that the landing image frame is obtained, the image frame representing the landing of the target object is captured after a period of time, the processor is triggered to start the stopping storage process of the captured image frame, and whether the moving target object appears again is analyzed in real time, and the motion analysis and landing position analysis process of the newly appeared target object are started. Of course, if a single training analysis is set in advance, after the processor is triggered to start the stopping storage process of the captured image frame, the target camera stops capturing the motion scene image of the object, i.e., it no longer analyzes whether the moving target object appears again in real time.

[0091] Preferably, in the embodiment of the present application, the target camera is installed in the area close to one end of the landing of the target object, and the height of the target camera is close to the ground. Further, the height of the target camera is 20 cm away from the ground, so that the target camera of the monocular vision system can more effectively capture the motion process of the target object in the motion scene.

[0092] Based on the inventive concept of the above-mentioned single-vision-based landing point automatic reporting method, the embodiment of the present application provides a single-vision-based landing point automatic reporting system, comprising:

[0093] A motion scene image acquisition module is configured to acquire a motion scene image of a thrown object in real time based on a target camera.

[0094] A target motion parameter analysis module is configured to analyze the motion of a target thrown object based on image frames acquired after a first time, wherein the first time is the time when the target thrown object starts to appear in the real-time acquired motion scene image frame.

[0095] A motion target image storage module is configured to trigger the processor to start a storage process of the acquired image frames at a second time, wherein the second time is the time when the motion position of the target thrown object meets a preset storage condition.

[0096] A landing position determination module is configured to analyze the landing position of the target thrown object based on the stored image frames.

[0097] It should be noted that the single-vision-based landing point automatic reporting system provided in the embodiment is only used as an example to illustrate the division of the above-mentioned functional modules when analyzing the landing position of a thrown object in a field event. In actual applications, the above-mentioned functions can be completed by different functional modules according to needs, i.e., the internal structure of the device is divided into different functional modules to complete all or part of the functions described above. In addition, the single-vision-based landing point automatic reporting system provided in the embodiment and the single-vision-based landing point automatic reporting method provided in the above-mentioned embodiment belong to the same concept, and the specific implementation process is described in detail in the method embodiment, which will not be repeated here.

[0098] The present application is not limited to the above-mentioned specific embodiments, and those skilled in the art can make various modifications based on the above-mentioned concept without creative labor, and all the modifications fall within the protection scope of the present application.

Claims

1. A method for automatically reporting landing points based on monocular vision, characterized in that: include: Based on the target camera, the moving scene image of the thrown object is collected in real time; analyzing the motion of the target thrown object based on image frames acquired after a first time, wherein the first time is the moment when the target thrown object begins to appear in the real-time acquired motion scene image frames; triggering the processor to start a storage process for the captured image frame at a second time, wherein the second time is the moment when the motion position of the target thrown object meets a preset storage condition; Analyzing the landing position of the target thrown object based on the stored image frames; The step of analyzing the landing point of the target thrown object based on the stored image frames includes: Based on the stored captured image frames, a landing image frame corresponding to the moment when the target thrown object lands is acquired; Identify and analyze the position of the target thrown object based on the landing image frame; The step of obtaining a landing image frame corresponding to the moment when the thrown object lands on the ground comprises: Based on the captured image frames stored up to the current moment, the position of the target thrown object on the captured image frames is obtained in chronological order; generating motion direction feature data of the target thrown object based on the position of the target thrown object in two adjacent image frames; Determining, based on the motion direction characteristic data, the correlation between the motion direction characteristic data before the current moment and the motion direction characteristic data between the current moment and the next moment; If the correlation is less than a preset correlation threshold, it is determined that the target thrown object has fallen to the ground, and the image frames collected at the current moment and the next moment are determined as falling image frames.

2. The method for automatically reporting a landing point based on monocular vision according to claim 1, characterized in that: Before the target camera is used to collect images of the motion scene of the thrown object in real time, the method includes: calibrating the target camera to determine the real coordinates corresponding to the positions of each pixel point in the image collected by the target camera in the motion scene.

3. The method for automatically reporting landing points based on monocular vision according to claim 1, characterized in that: The method for obtaining the first time includes: using a frame difference method to detect whether a target thrown object appears in the motion scene image collected at the current moment based on the motion scene image collected at the current moment and the motion scene image collected at the previous moment. If so, the current moment is the first time.

4. The method for automatically reporting a landing point based on monocular vision according to claim 1, characterized in that: The analyzing the motion of the target thrown object based on the image frames collected after the first time includes: Based on the image frames acquired after the first time, performing a frame difference analysis on a selected area in the image frames to determine whether a target thrown object appears in the selected area; The triggering of the processor to start the storage process of the acquired image frame at the second time includes: When a target thrown object appears in the selected area in the image frame, the processor is triggered to start a storage process for the acquired image frame.

5. The method for automatically reporting a landing point based on monocular vision according to claim 4, characterized in that: The selected area is an area consisting of multiple rows of pixels in the image frame.

6. The method for automatically reporting landing points based on monocular vision according to claim 1, characterized in that: The method of performing frame difference analysis on a selected area in the image frame based on the image frame acquired after the first time to determine whether the target thrown object appears in the selected area includes the following steps: Analyzing initial motion parameters of the target object at the first time of being thrown, and predicting a theoretical landing time of the target object based on the initial motion parameters, wherein the initial motion parameters include initial motion displacement, initial velocity magnitude, and initial velocity direction; The target area where the thrown object is located during a preset time period before the predicted value of the theoretical landing time is used as the selected area; Based on the image frames acquired after the first time, a frame difference analysis is performed on a selected area in the image frames to determine whether a target thrown object appears in the selected area.

7. The method for automatically reporting a landing point based on monocular vision according to claim 6, characterized in that: The preset duration is determined according to the size of the storage resources occupied by the target camera's captured images per unit time.

8. The method for automatically reporting a landing point based on monocular vision according to claim 1, characterized in that: The identifying and analyzing the position of the target thrown object based on the landing image frame includes: Determining that the target thrown object has not experienced trajectory interruption based on motion direction feature data generated based on the actual target thrown object position at the next moment and the current moment; Obtaining a first position of the target thrown object represented by the image frame at the current moment and a second position represented by the image frame at the next moment in the landing image frame; The first position is determined to be a landing point or a landing point between the first position and a second position based on the altitude data of the first position.

9. The method for automatically reporting landing points based on monocular vision according to claim 1, characterized in that: The target camera is installed in an area close to the end where the target object lands, and is installed at a height close to the ground.

10. The method for automatically reporting landing points based on monocular vision according to claim 1, characterized in that: The target camera is installed at a height of 20 cm from the ground.

11. A landing point automatic reporting system based on monocular vision, characterized in that: include: A motion scene image acquisition module is used to acquire motion scene images of thrown objects in real time based on a target camera; a target motion parameter analysis module configured to analyze the motion of a target thrown object based on image frames acquired after a first time, wherein the first time is the moment when the target thrown object begins to appear in the real-time acquired motion scene image frames; A moving target image storage module is used to trigger the processor to start a storage process for the captured image frame at a second time, where the second time is the moment when the moving position of the target thrown object meets a preset storage condition; A landing position determination module is used to analyze the landing position of the target thrown object based on the stored image frames; The step of analyzing the landing point of the target thrown object based on the stored image frames includes: Based on the stored captured image frames, a landing image frame corresponding to the moment when the target thrown object lands is acquired; Identify and analyze the position of the target thrown object based on the landing image frame; The step of obtaining a landing image frame corresponding to the moment when the thrown object lands on the ground comprises: Based on the captured image frames stored up to the current moment, the position of the target thrown object on the captured image frames is obtained in chronological order; generating motion direction feature data of the target thrown object based on the position of the target thrown object in two adjacent image frames; Determining, based on the motion direction characteristic data, the correlation between the motion direction characteristic data before the current moment and the motion direction characteristic data between the current moment and the next moment; If the correlation is less than a preset correlation threshold, it is determined that the target thrown object has fallen to the ground, and the image frames collected at the current moment and the next moment are determined as falling image frames.

Images