Flying platform-based moving target remote identification method and storage medium

Abstract

The invention discloses a moving target remote identification method based on a flight platform and a storage medium. The method comprises the following steps: detecting a target in an image by using a target detection network; performing pipeline association on the target detected by each frame of image, and updating the position of the target in the pipeline; determining a key frame image, mapping each subsequent frame image with the key frame image, and calculating a perspective transformation matrix of each subsequent frame image relative to the key frame image; mapping the central point of each detection frame of each subsequent frame image to the key frame image according to the calculated perspective transformation matrix; comparing the target mapping result of the current frame image in the same pipeline with the target mapping result of the previous frame image of the current frame image according to the pipeline association result, and confirming the motion state of the target; and setting the pipeline of the moving target to be in a moving state, setting the pipeline of the static target to be in a static state, outputting the position of the pipeline target which is finally determined to be in the moving state, and identifying and tracking the moving target.

Description

Technical Field

[0001] The invention relates to the field of image recognition and tracking, and in particular to a moving target long-distance recognition method based on a flying platform and a storage medium. Background Art

[0002] In military reconnaissance, search and rescue, or environmental monitoring missions, quickly and accurately locating the target is often the key to mission success. For moving targets, the target is in a moving state. If the target cannot be identified and tracked in time, it will move out of the field of view, resulting in mission failure. In addition, long-distance identification and tracking of targets can also help improve the real-time and accuracy of information acquisition.

[0003] The existing target detection technology based on the target detection network uses the information of a single frame image to detect the trained targets in the scene, but it cannot determine the motion state of the target and cannot distinguish between moving targets and stationary targets. In addition, this method only uses the information of the target single frame image, and has poor reliability. It recognizes the target by accumulating confidence, and the recognition speed is slow and the distance is short.

[0004] Existing motion target detection methods based on optical flow, frame difference, and background modeling have large computational complexity and long time consumption, which are not conducive to scenarios with high real-time requirements. In addition, on a stationary platform, this type of motion target detection effect may be better, but for a flying platform, due to the movement of the platform itself, the static background is often detected as a moving target, and the false alarm rate is high. Summary of the invention

[0005] In view of the technical defects and technical drawbacks in the prior art, the embodiments of the present invention provide a method and storage medium for long-distance identification of moving targets based on a flying platform to overcome the above problems or at least partially solve the above problems. The specific scheme is as follows:

[0006] As a first aspect of the present invention, a method for long-distance identification of moving targets based on a flying platform is provided, the method comprising:

[0007] When the flying platform collects images containing targets and scenes, the target detection network is used to detect the targets in the images;

[0008] Perform pipeline association for the targets detected in each frame of image and update the position of the targets in the pipeline;

[0009] Determine the key frame image, and map each subsequent frame image to the key frame image, and calculate the perspective transformation matrix of each subsequent frame image relative to the key frame image;

[0010] According to the calculated perspective transformation matrix, the center point of each detection box of each subsequent frame image is mapped to the key frame image;

[0011] According to the result of pipeline association, the result of target mapping of the current frame image in the same pipeline is compared with the result of target mapping of the frame image before the current frame image to confirm the target motion state;

[0012] The pipeline of the moving target is set to a moving state, the pipeline of the stationary target is set to a stationary state, and the position of the pipeline target finally determined to be in a moving state is output to identify and track the moving target.

[0013] Furthermore, when the flying platform collects an image containing a target and a scene, detecting the target in the image using a target detection network includes:

[0014] Step 101, collect images containing targets and scenes, mark the targets, and save the images to an established target image database;

[0015] Step 102, inputting the images in the target image database into the target detection network for training, and after the target detection network is trained, converting the target detection network into a model format supported by the ARM hardware platform;

[0016] Step 103, loading the target detection model into the ARM hardware platform, performing inference on the ARM hardware platform, and detecting all targets in the current frame image.

[0017] Furthermore, the step of associating the target detected in each frame of the image with the pipeline and updating the position of the target in the pipeline includes:

[0018] Step 201, for the first frame image detected, initialize a pipeline for each detection result, and use the position of the target detection in each pipeline to initialize a tracking chain;

[0019] Step 202: In each subsequent frame image, first determine whether the number of target pipes in the previous frame image is greater than 0. If the number of target pipes is greater than 0, use the tracking chain in each pipe to predict the position of the target in the corresponding pipe in the current frame image.

[0020] Step 203, calculating the IOU value between the predicted box of the target in the current frame image and all the detection results in each pipeline;

[0021] Step 204: According to the IOU value, Hungarian matching is used to obtain the matching relationship between each pipeline target prediction box and all detection boxes, and the position and tracking chain of the target in the pipeline are updated.

[0022] Further, step 204 includes:

[0023] a. If the detection box does not match all the target pipelines, a pipeline is initialized for the detection box, and a tracking chain is initialized based on the detection results;

[0024] b. If the detection frame successfully matches a target pipeline, the detection result is used to update the position and tracking chain of the target in the pipeline;

[0025] c. If the target pipeline does not match all detection frames, the predicted result is used to update the position of the target in the pipeline and mark the pipeline as not matching the current frame image.

[0026] Furthermore, in step 202, if the number of pipelines is not greater than 0, the operation in step 201 is repeated to initialize a pipeline for each detection result.

[0027] Furthermore, the determining of the key frame image, mapping each subsequent frame image to the key frame image, and calculating the perspective transformation matrix of each subsequent frame image relative to the key frame image specifically includes:

[0028] Step 301, in the first frame image detected, the image is divided into four quadrants with the center point as the boundary, and an image block of a preset size is randomly selected in each of the four quadrants of the image to perform tracking initialization, and the frame image is the key frame image;

[0029] Step 302, in each subsequent frame image, track the four image blocks selected from the key frame image, and use the center points of the four tracking frames of the current frame image and the center points of the four tracking frames in the key frame image to calculate the perspective transformation matrix of the current frame image and the key frame image.

[0030] Furthermore, the method further includes: determining whether it is necessary to update the key frame image. There are three conditions for updating the key frame image. If any one of the following three conditions is met, the key frame image needs to be updated:

[0031] Condition 1: The tracking confidence of the four image blocks of the current frame image is lower than the confidence threshold;

[0032] Condition 2: The tracking positions of the four image blocks in the current frame image are at the edge of the image;

[0033] Condition 3: The maximum IOU value between the tracking boxes of the four image blocks of the current frame image and all the detection boxes of the current frame image is greater than 0.

[0034] Furthermore, mapping the center point of each detection frame of each subsequent frame image to the key frame image according to the calculated perspective transformation matrix includes:

[0035] If the frame image is a key frame image, no mapping is required, and the coordinates of the target mapping point in the pipeline are updated using the coordinates of the center point of the detection frame;

[0036] If the current frame image is not a key frame image, the center point of the detection box is calculated, and the coordinates of the center point mapped to the key frame image are calculated using the perspective transformation matrix, and the coordinates of the target mapping point in the pipeline are updated.

[0037] Furthermore, the step of comparing the result of the target mapping of the current frame image in the same pipeline with the result of the target mapping of the frame image before the current frame image according to the result of the pipeline association to confirm the target motion state includes:

[0038] Step 501, assuming that the coordinates of a pipeline target mapping point in the current frame image are (x1, y1), and the coordinates of the corresponding pipeline target mapping point in the previous frame image are (x2, y2), if x1-x2>0, the target current frame image moves in the positive x direction, the x positive direction movement count is increased by 1, the x reverse direction movement count is 0, and the x direction movement displacement is (x1-x2); if y1-y2>0, the y positive direction movement count is increased by 1, the y reverse direction movement count is 0, and the y direction movement displacement is (y1-y2).

[0039] Step 502, within 1 second, the forward and reverse movement of the target x is counted cumulatively, and the displacement of the movement in the x direction is cumulatively added; the forward and reverse movement of the target y is counted cumulatively, and the displacement of the movement in the y direction is added;

[0040] Step 503, if the accumulated count of the x-direction movement is greater than the x-direction calculation threshold and the accumulated count of the x-direction movement is 0, and the x-direction movement displacement is greater than the x-direction displacement threshold, the target moves in the x-direction, and the x-direction movement flag is set to 1;

[0041] If the accumulated count of the x-reverse motion is greater than the x-reverse count threshold and the accumulated count of the x-positive motion is 0, and the x-direction motion displacement is greater than the x-direction displacement threshold, the target moves in the x-reverse direction, and the x-direction motion flag is set to 1;

[0042] If the accumulated count of the y-direction movement is greater than the y-direction calculation threshold and the accumulated count of the y-direction movement is 0, and the y-direction movement displacement is greater than the y-direction displacement threshold, the target moves in the y-direction, and the y-direction movement flag is set to 1;

[0043] If the accumulated count of the y-reverse motion is greater than the y-reverse calculation threshold and the accumulated count of the y-positive motion is 0, and the y-direction motion displacement is greater than the y-direction displacement threshold, the target moves in the y-reverse direction, and the y-direction motion flag is set to 1;

[0044] Step 504: If the x-direction motion flag is set to 1 or the y-direction motion flag is set to 1 in step 503, the corresponding target is considered to be a moving target, otherwise it is considered to be a stationary target.

[0045] As a second aspect of the present invention, a computer-readable storage medium is provided, in which a computer program is stored. When the computer program is executed by a computer, the computer executes any of the above-mentioned methods for long-range identification of moving targets based on a flying platform.

[0046] The present invention has the following beneficial effects:

[0047] (1) In the task of identifying and tracking moving targets, in multi-target scenes, moving targets can be quickly and accurately identified and tracked.

[0048] (2) It can realize ultra-long-distance recognition of moving targets and prevent the targets from moving out of the field of view.

[0049] (3) When performing registration of non-key frame images and key frame images, the present invention adopts a method of tracking four image blocks in the background for registration, thereby avoiding the complex process of conventional feature point matching and effectively reducing the time consumption of moving target detection.

[0050] (4) This technology utilizes the temporal and spatial continuity of the target in the image sequence to effectively improve the recall rate and accuracy of moving target detection. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] Figure 1 A flow chart of a method for long-distance identification of moving targets based on a flying platform provided in an embodiment of the present invention;

[0052] Figure 2 A target detection flow chart provided by an embodiment of the present invention;

[0053] Figure 3 A target pipeline association flow chart provided for an embodiment of the present invention;

[0054] Figure 4 An image registration flow chart provided by an embodiment of the present invention;

[0055] Figure 5 A flow chart of target motion attribute recognition provided for an embodiment of the invention. DETAILED DESCRIPTION

[0056] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0057] See also Figure 1, a method for long-distance identification of moving targets based on a flying platform is provided in an embodiment of the present invention, and the method comprises:

[0058] Step 1: When the flying platform collects an image containing a target and a scene, the target detection network is used to detect the target in the image;

[0059] Step 2: Associate the targets detected in each frame of the image with the pipeline and update the position of the targets in the pipeline;

[0060] Step 3, determine the key frame image, and map each subsequent frame image to the key frame image, and calculate the perspective transformation matrix of each subsequent frame image relative to the key frame image;

[0061] This step is used to solve the influence caused by the movement of the platform. Optionally, the present invention selects four areas in the key frame image for tracking initialization, and each subsequent frame image tracks the four areas respectively, and uses the center point positions of the four area tracking to calculate the perspective transformation matrix of each subsequent frame image relative to the key frame image;

[0062] Step 4: Map the center point of each detection frame of each subsequent frame image to the key frame image according to the calculated perspective transformation matrix, wherein the center point of the detection frame is mapped to the coordinates in the key frame image, i.e., the coordinates of the target mapping point;

[0063] Step 5, according to the result of pipeline association, the result of target mapping of the current frame image in the same pipeline is compared with the result of target mapping corresponding to one or more frames before the current frame image to confirm the target motion state;

[0064] For example, the result of the target mapping of the current frame image in the same pipeline is compared with the result of the target mapping of the previous frame image, and the movement direction and displacement of the target of the current frame image relative to the target of the previous frame image are counted. After accumulating multiple frames of images, if the target moves a certain displacement in a certain direction for multiple consecutive frames of images, the target is considered to be a moving target, otherwise the target is a stationary target.

[0065] Step 6: Set the pipeline of the moving target to a moving state, set the pipeline of the stationary target to a stationary state, and output the target position of the pipeline finally determined to be in a moving state, and identify and track the moving target.

[0066] The present invention performs pipeline association through targets detected by images. In the task of identifying and tracking moving targets, for multi-target scenes, the moving targets can be quickly and accurately identified and accurately tracked. When performing registration of non-key frame images with key frame images, the present invention adopts a method of tracking four image blocks in the background for registration, avoiding the complex process of conventional feature point matching and effectively reducing the time consumption of moving target detection.

[0067] See also Figure 2 As a specific embodiment of the present invention, the pipeline association of the target detected in each frame image and updating the position of the target in the pipeline specifically includes:

[0068] Step 101, collect images containing targets and scenes, mark the targets, and save the images to an established target image database;

[0069] Step 102, inputting the images in the target image database into the target detection network for training, and after the target detection network is trained, converting the target detection network into a model format supported by the ARM hardware platform;

[0070] Step 103, loading the target detection model into the ARM hardware platform, performing inference on the ARM hardware platform, and detecting all targets in the current frame image.

[0071] See also Figure 3 As a specific embodiment of the present invention, the pipeline association of the target detected in each frame image and updating the position of the target in the pipeline specifically includes:

[0072] Step 201, for the first frame image detected, initialize a pipeline for each detected target, and initialize a tracking chain using the position of the target detection in each pipeline.

[0073] Step 202: In each subsequent frame image, first determine whether the number of target pipelines in the previous frame image is greater than 0. If the number of target pipelines is greater than 0, use the tracking chain in each pipeline to predict the position of the target in the corresponding pipeline in the current frame image; if the number of pipelines is not greater than 0, repeat the operation in step 201 to initialize a pipeline for each detection target.

[0074] Step 203, calculate the IOU value between the predicted box of the target in the current frame image and all the detection boxes in each pipeline.

[0075] Step 204, based on the IOU value, Hungarian matching is used to obtain the matching relationship between each pipeline target prediction box and all detection boxes. Three matching relationships will appear, as follows:

[0076] a) If the detection box does not match all the target pipelines, a pipeline is initialized for the detection box, and a tracking chain is initialized based on the detection results;

[0077] b) If the detection frame successfully matches a target pipeline, the detection result is used to update the position and tracking chain of the target in the pipeline;

[0078] c) If the target pipeline does not match all detection frames, the predicted result is used to update the position of the target in the pipeline and mark the pipeline as not matching the current frame image.

[0079] Step 205: Pipeline association is completed.

[0080] See also Figure 4 As a specific embodiment of the present invention, the determining of the key frame image, mapping each subsequent frame image with the key frame image, and calculating the perspective transformation matrix of each subsequent frame image relative to the key frame image specifically include:

[0081] Step 301, in the first frame image detected, the image is divided into four quadrants with the center point of the image as the boundary, and a 20*15 image block is randomly selected from each of the four quadrants of the image to perform tracking initialization. The frame image is the key frame image, in which each image block is tracked by a tracking frame;

[0082] Step 302, in each subsequent frame image, first track the four image blocks selected in the key frame image, and use the center points of the four tracking frames of the current frame image and the center points of the four tracking frames in the key frame image to calculate the perspective transformation matrix of the current frame image and the key frame image.

[0083] Optionally, the method further includes: determining whether the next frame image needs to update the key frame image, and there are three conditions for updating the key frame image. As long as any one of the following three conditions is met, the key frame image needs to be updated:

[0084] Condition 1: The tracking confidence of the four image blocks of the current frame image is lower than the confidence threshold;

[0085] Condition 2: The tracking positions of the four image blocks in the current frame image are at the edge of the image;

[0086] Condition 3: The maximum IOU value between the tracking boxes of the four image blocks of the current frame image and all the detection boxes of the current frame image is greater than 0.

[0087] As long as any one of the above three conditions is met, the key frame image needs to be updated, and the current frame image or the frame image previous to the current frame image is updated to the key frame image according to the method of step 301 .

[0088] As a specific embodiment of the present invention, the pipeline association of the target detected in each frame image and updating the position of the target in the pipeline specifically includes:

[0089] Step 401: if the frame image is a key frame image, no mapping is required, and the coordinates of the target mapping point in the pipeline are updated using the coordinates of the center point of the detection frame;

[0090] Step 402: If the current frame image is not a key frame image, the center point of the detection frame is calculated, and the coordinates of the center point mapped to the key frame image are calculated using the perspective transformation matrix, and the coordinates of the target mapping point in the pipeline are updated.

[0091] See also Figure 5 As a specific embodiment of the present invention, the result of the target mapping of the current frame image in the same pipeline is compared with the result of the target mapping of the frame image before the current frame image according to the result of the pipeline association, and the target motion state is confirmed, including:

[0092] Step 501, assuming that the coordinates of a pipeline target mapping point in the current frame image are (x1, y1), and the coordinates of the corresponding pipeline target mapping point in the previous frame image are (x2, y2), if x1-x2>0, the target moves in the positive x direction in the current frame image, the x positive direction movement count is increased by 1, the x reverse direction movement count is 0, and the x direction movement displacement is (x1-x2); if y1-y2>0, the y positive direction movement count is increased by 1, the y reverse direction movement count is 0, and the y direction movement displacement is (y1-y2).

[0093] Step 502, within 1 second, the forward and reverse movement of the target x is counted cumulatively, and the displacement of the movement in the x direction is cumulatively added; the forward and reverse movement of the target y is counted cumulatively, and the displacement of the movement in the y direction is added;

[0094] Step 503, if the accumulated count of the x-direction movement is greater than the x-direction calculation threshold and the accumulated count of the x-direction movement is 0, and the x-direction movement displacement is greater than the x-direction displacement threshold, the target moves in the x-direction, and the x-direction movement flag is set to 1;

[0095] If the accumulated count of the x-reverse motion is greater than the x-reverse count threshold and the accumulated count of the x-positive motion is 0, and the x-direction motion displacement is greater than the x-direction displacement threshold, the target moves in the x-reverse direction, and the x-direction motion flag is set to 1;

[0096] If the accumulated count of the y-direction movement is greater than the y-direction calculation threshold and the accumulated count of the y-direction movement is 0, and the y-direction movement displacement is greater than the y-direction displacement threshold, the target moves in the y-direction, and the y-direction movement flag is set to 1;

[0097] If the accumulated count of the y-reverse motion is greater than the y-reverse calculation threshold and the accumulated count of the y-positive motion is 0, and the y-direction motion displacement is greater than the y-direction displacement threshold, the target moves in the y-reverse direction, and the y-direction motion flag is set to 1;

[0098] Step 504, if the x-direction motion flag is set to 1 or the y-direction motion flag is set to 1 in step 503, the target is considered to be a moving target and the pipe where the target is located is set to a moving pipe; otherwise, the target is considered to be a stationary target and the pipe where the target is located is set to a stationary pipe.

[0099] As another embodiment of the present invention, a computer-readable storage medium is also provided, in which a computer program is stored. When the computer program is executed by a computer, the computer executes any of the above-mentioned methods for long-distance identification of moving targets based on a flying platform.

[0100] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for long-distance identification of moving targets based on a flying platform, characterized in that: The method comprises: When the flying platform collects images containing targets and scenes, the target detection network is used to detect the targets in the images; Perform pipeline association for the targets detected in each frame of image and update the position of the targets in the pipeline; Determine the key frame image, and map each subsequent frame image to the key frame image, and calculate the perspective transformation matrix of each subsequent frame image relative to the key frame image; According to the calculated perspective transformation matrix, the center point of each detection box of each subsequent frame image is mapped to the key frame image; According to the result of pipeline association, the result of target mapping of the current frame image in the same pipeline is compared with the result of target mapping of the frame image before the current frame image to confirm the target motion state; The pipeline of the moving target is set to a moving state, the pipeline of the stationary target is set to a stationary state, and the position of the pipeline target finally determined to be in a moving state is output to identify and track the moving target.

2. The method for long-distance identification of moving targets based on a flying platform according to claim 1 is characterized in that: When the flying platform collects an image containing a target and a scene, detecting the target in the image using a target detection network includes: Step 101, collect images containing targets and scenes, mark the targets, and save the images to an established target image database; Step 102, inputting the images in the target image database into the target detection network for training, and after the target detection network is trained, converting the target detection network into a model format supported by the ARM hardware platform; Step 103, loading the target detection model into the ARM hardware platform, performing inference on the ARM hardware platform, and detecting all targets in the current frame image.

3. The method for long-distance identification of moving targets based on a flying platform according to claim 1, characterized in that: The pipeline association of the target detected in each frame image and updating the position of the target in the pipeline includes: Step 201, for the first frame image detected, initialize a pipeline for each detection result, and use the position of the target detection in each pipeline to initialize a tracking chain; Step 202: In each subsequent frame image, first determine whether the number of target pipes in the previous frame image is greater than 0. If the number of target pipes is greater than 0, use the tracking chain in each pipe to predict the position of the target in the corresponding pipe in the current frame image. Step 203, calculating the IOU value between the prediction box of the target in the current frame image and all the detection boxes in each pipeline; Step 204: According to the IOU value, Hungarian matching is used to obtain the matching relationship between each pipeline target prediction box and all detection boxes, and the position and tracking chain of the target in the pipeline are updated.

4. The method for long-distance identification of moving targets based on a flying platform according to claim 3 is characterized in that: Step 204 includes: a. If the detection box does not match all the target pipelines, a pipeline is initialized for the detection box, and a tracking chain is initialized based on the detection results; b. If the detection frame successfully matches a target pipeline, the detection result is used to update the position and tracking chain of the target in the pipeline; c. If the target pipeline does not match all detection frames, the predicted result is used to update the position of the target in the pipeline and mark the pipeline as not matching the current frame image.

5. The method for long-distance identification of moving targets based on a flying platform according to claim 3 is characterized in that: In step 202, if the number of pipelines is not greater than 0, the operation in step 201 is repeated to initialize a pipeline for each detection result.

6. The method for long-distance identification of moving targets based on a flying platform according to claim 1, characterized in that: The determining of the key frame image, mapping each subsequent frame image to the key frame image, and calculating the perspective transformation matrix of each subsequent frame image relative to the key frame image specifically includes: Step 301, in the first frame image detected, the image is divided into four quadrants with the center point of the image as the boundary, and an image block of a preset size is randomly selected in each of the four quadrants of the image to perform tracking initialization, and this frame image is used as a key frame image; Step 302, in each subsequent frame image, track the four image blocks selected in the key frame image, and use the center points of the four tracking frames of the current frame image and the center points of the four tracking frames in the key frame image to calculate the perspective transformation matrix of the current frame image and the key frame image.

7. The method for long-distance identification of moving targets based on a flying platform according to claim 6 is characterized in that: The method further includes: determining whether it is necessary to update the key frame image. There are three conditions for updating the key frame image. If any one of the following three conditions is met, the key frame image needs to be updated: Condition 1: The tracking confidence of the four image blocks of the current frame image is lower than the confidence threshold; Condition 2: The tracking positions of the four image blocks in the current frame image are at the edge of the image; Condition 3: The maximum IOU value between the tracking boxes of the four image blocks of the current frame image and all the detection boxes of the current frame image is greater than 0.

8. The method for long-distance identification of moving targets based on a flying platform according to claim 1, characterized in that: Mapping the center point of each detection frame of each subsequent frame image to the key frame image according to the calculated perspective transformation matrix includes: If the frame image is a key frame image, no mapping is required, and the coordinates of the target mapping point in the pipeline are updated using the coordinates of the center point of the detection frame; If the current frame image is not a key frame image, the center point of the detection box is calculated, and the coordinates of the center point mapped to the key frame image are calculated using the perspective transformation matrix, and the coordinates of the target mapping point in the pipeline are updated.

9. The method for long-distance identification of moving targets based on a flying platform according to claim 1, characterized in that: The step of comparing the target mapping result of the current frame image in the same pipeline with the target mapping result of the frame image before the current frame image according to the result of the pipeline association to confirm the target motion state includes: Step 501, assuming that the coordinates of a certain pipeline target mapping point of the current frame image are (x1, y1), and the coordinates of the corresponding pipeline target mapping point of the previous frame image are (x2, y2), if x1-x2>0, then the target current frame image moves in the positive x direction, the positive x direction movement count is increased by 1, the negative x direction movement count is 0, and the x direction movement displacement is (x1-x2); if y1-y2>0, the positive y direction movement count is increased by 1, the negative y direction movement count is 0, and the y direction movement displacement is (y1-y2); Step 502, within 1 second, the forward and reverse movement of the target x is counted cumulatively, and the displacement of the movement in the x direction is cumulatively added; the forward and reverse movement of the target y is counted cumulatively, and the displacement of the movement in the y direction is added; Step 503, if the accumulated count of the x-direction movement is greater than the x-direction calculation threshold and the accumulated count of the x-direction movement is 0, and the x-direction movement displacement is greater than the x-direction displacement threshold, the target moves in the x-direction, and the x-direction movement flag is set to 1; If the accumulated count of the x-reverse motion is greater than the x-reverse count threshold and the accumulated count of the x-positive motion is 0, and the x-direction motion displacement is greater than the x-direction displacement threshold, the target moves in the x-reverse direction, and the x-direction motion flag is set to 1; If the accumulated count of the y-direction movement is greater than the y-direction calculation threshold and the accumulated count of the y-direction movement is 0, and the y-direction movement displacement is greater than the y-direction displacement threshold, the target moves in the y-direction, and the y-direction movement flag is set to 1; If the accumulated count of the y-reverse motion is greater than the y-reverse calculation threshold and the accumulated count of the y-positive motion is 0, and the y-direction motion displacement is greater than the y-direction displacement threshold, the target moves in the y-reverse direction, and the y-direction motion flag is set to 1; Step 504: If the x-direction motion flag is set to 1 or the y-direction motion flag is set to 1 in step 503, the corresponding target is considered to be a moving target, otherwise it is considered to be a stationary target.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, and when the computer program is executed by a computer, the computer executes the method for long-distance identification of moving targets based on a flying platform as described in any one of claims 1 to 9.

Images