A method for detecting and tracking objects thrown from high altitude

Abstract

The present invention discloses a method for detecting and tracking high-altitude objects, which includes obtaining a binary image corresponding to a surveillance video frame image; obtaining a plurality of rectangular blocks; establishing a candidate object based on the rectangular blocks, updating a matching pool and a matching forest; determining a trajectory tree to be deleted; saving a suspected parabolic path, removing all the trajectory trees to be deleted from the matching forest; and determining a parabolic event. The method for detecting and tracking high-altitude objects of the present invention determines and filters the suspected parabolic path of the tracking target through a trajectory tree, thereby finding a confirmed parabolic event and effectively reducing the probability of missed reports and false reports.

Description

Technical Field

[0001] The present invention relates to the field of visual monitoring technology, and more specifically, to a method for detecting and tracking objects thrown from a high altitude. Background Art

[0002] Due to the rapid development of domestic infrastructure technology, the building height of domestic buildings is constantly being refreshed, and the density of buildings in urban construction is also constantly increasing, which has also caused the emergence of the safety hazard of high-altitude objects. In the past, when the problem of high-altitude objects was not taken seriously, even if someone was injured by high-altitude objects, it was difficult to trace the cause and origin of the incident, which caused great trouble to the public's life safety. In the existing technology, cameras for real-time monitoring are usually installed in the monitoring area downstairs, and video surveillance is performed through the camera. When a problem occurs, the video is traced back to query the origin of the incident.

[0003] When existing technologies are performing trajectory tracking, there is a problem that when the detection model accidentally fails to successfully detect the target, the Hungarian algorithm will not be able to successfully match between frames, resulting in missed recognition; or the screening of false alarms uses a simple downward acceleration rule, which cannot exclude a large number of flying birds, dust, and flying insects, and will miss dangerous objects but those that are not obviously accelerated, such as cigarette butts.

[0004] Therefore, it is urgent to propose a high-altitude dropped object detection and tracking method to solve the problem raised. Summary of the invention

[0005] 1. Technical issues to be resolved

[0006] Based on this, the present invention provides a method for detecting and tracking high-altitude parabolic objects, which adopts a block method to control the influence of high-frequency noise in the image, cooperates with a trajectory tree to determine and filter the object trajectory, finds the determined parabolic trajectory, and effectively reduces missed reports and false alarms.

[0007] (II) Technical solution

[0008] In order to solve the above technical problems, the present invention proposes a method for detecting and tracking objects thrown from a high altitude, which comprises the following steps:

[0009] Step S100, obtaining a binary image corresponding to a surveillance video frame image;

[0010] Step S200, obtaining a plurality of rectangular blocks based on bright blocks in the binary image;

[0011] Step S300, establishing a set of objects to be selected based on the rectangular blocks in the current frame binary image, performing matching calculations on the objects to be selected and the nodes in the matching pool one by one, generating new trajectory tree nodes according to the calculation results, and then updating the matching pool and the matching forest; wherein the number of nodes in the matching pool is initially empty, the matching forest is composed of trajectory trees, the number of trajectory trees in the matching forest is initially empty, and the trajectory tree is composed of trajectory tree nodes connected in a tree-like manner;

[0012] Step S400, determining whether the trajectory tree meets the preset to-be-deleted condition, if the trajectory tree meets the preset to-be-deleted condition, determining the trajectory tree as a trajectory tree to be deleted;

[0013] Step S500, obtaining all trajectory paths corresponding to the trajectory tree to be deleted, determining whether there is a trajectory path in the trajectory path corresponding to the trajectory tree to be deleted that satisfies a preset suspected parabolic condition, and if there is a trajectory path that satisfies the preset suspected parabolic condition, determining the trajectory path to be a suspected parabolic path, and saving the suspected parabolic path to obtain a saved suspected parabolic path set, and then removing all trajectory trees to be deleted from the matching forest;

[0014] Step S600: determine whether the stored suspected parabolic path meets the preset parabolic condition. If there is a suspected parabolic path that meets the preset parabolic condition, determine that the suspected parabolic path is a parabolic event.

[0015] (III) Beneficial effects

[0016] Compared with the prior art, the present invention provides a method for detecting and tracking high-altitude parabolic objects, which uses a trajectory tree to determine and filter the suspected parabolic paths of the tracking target, thereby finding out the confirmed parabolic events and effectively reducing the probability of missed reports and false alarms. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] The features and advantages of the present invention will be more clearly understood by referring to the accompanying drawings, which are schematic and should not be construed as limiting the present invention in any way. In the accompanying drawings:

[0018] Figure 1 It is a flow chart of a first method for detecting and tracking high-altitude objects disclosed in a preferred embodiment of the present invention;

[0019] Figure 2 It is a flow chart of a second method for detecting and tracking high-altitude objects disclosed in a preferred embodiment of the present invention;

[0020] Figure 3 It is a flow chart of a third method for detecting and tracking high-altitude objects disclosed in a preferred embodiment of the present invention;

[0021] Figure 4 is the image data corresponding to one of the PSNR result matrices disclosed in the preferred embodiment of the present invention;

[0022] Figure 5 yes Figure 4 A binary image corresponding to a PSNR result matrix data disclosed in the preferred embodiment of the present invention shown;

[0023] Figure 6 It is a state diagram of one of the trajectory trees disclosed in the preferred embodiment of the present invention. DETAILED DESCRIPTION

[0024] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the specific embodiments of the present invention are described in detail below in conjunction with the accompanying drawings. In the following description, many specific details are set forth to facilitate a full understanding of the present invention. However, the present invention can be implemented in many other ways different from those described herein, and those skilled in the art can make similar improvements without violating the connotation of the present invention, so the present invention is not limited by the specific implementation disclosed below.

[0025] Figure 1 is a flow chart of a first method for detecting and tracking high-altitude objects provided by an embodiment of the present invention, such as Figure 1 As shown, the method for detecting and tracking high-altitude objects includes steps S100 to S600, which are specifically as follows:

[0026] Step S100: Obtain a binary image corresponding to a surveillance video frame image, wherein bright spots in the binary image are proposed to be used as moving pixels in the frame image.

[0027] The method of step S100 comprises the following steps:

[0028] Step S110, obtaining frame images of surveillance video; a surveillance camera can be set up under a high-rise building to shoot surveillance video in real time in the direction of the floors, and each frame image of the surveillance video shot in real time by the surveillance camera is processed by the high-altitude object detection and tracking method of the present invention to monitor high-altitude objects in real time.

[0029] Step S120, obtaining a differential image corresponding to the surveillance video frame image; obtaining the differential image corresponding to the surveillance video frame image by a three-frame differential method, wherein the specific process of obtaining the differential image by the three-frame differential method is as follows: obtaining two differential images D(0) = |I(t)-I(t-1)| and D(1) = |I(t+1)-I(t)| from three consecutive frame images I(t-1), I(t), and I(t+1), and performing an intersection operation on the two differential images D(0) and D(1) respectively after binarization to obtain the target in the frame image at time t. The outline of the foreground target can be found and marked by connected domain analysis, and the bounding box of the target can be obtained. This process requires setting the binarization threshold of the frame difference image to extract the target outline, and setting the target size threshold to exclude some interfering targets, such as fallen leaves, flying birds, garbage, etc. moving near the camera; or obtaining the differential map corresponding to the frame image of the surveillance video through the inter-frame difference method, wherein the specific process of obtaining the differential map through the inter-frame difference method is as follows: the frame image information of the previous frame and the frame image information of the next frame are respectively subjected to differential processing to obtain the differential map.

[0030] Step S130, divide the difference image into several image blocks, obtain the PSNR result corresponding to each image block, use the PSNR result as the core calculation indicator to find the picture changes, so as to facilitate the change point detection of the frame image of the surveillance video, and use the block method to control the influence of high-frequency noise in the image, so as to effectively reduce the amount of calculation.

[0031] The method for obtaining the PSNR result corresponding to each image block in step S130 is specifically performed as follows:

[0032]

[0033]

[0034] Get the PSNR results corresponding to each image block, where I MAX Indicates the maximum intensity of the gray value of the pixel. For an 8-bit image, I MAX The value is 255; (u,v) represents the block coordinates of the image block in the difference image, u is the horizontal coordinate, and v is the vertical coordinate; (i,j) represents the pixel coordinates of the pixel point in a single image block, i is the horizontal coordinate, and j is the vertical coordinate; c represents the color channel, and its value is 3 for RGB images; t represents the time sequence number corresponding to the difference image, C represents the number of channels, and B represents the side length of the image block.

[0035] Step S140: Arrange the PSNR results corresponding to each image block in a matrix manner to form a PSNR result matrix.

[0036] In this embodiment, taking the frame image of the 1920*1080 surveillance video as an example, a 1920*1080 difference image is obtained by the three-frame difference method, and a 4*4 pixel image block is used to calculate the PSNR result corresponding to each image block, and a 480*270 floating-point matrix composed of multiple PSNR results as elements is obtained, such as Figure 4 As shown, the dark spots are the parts with more obvious changes in the original surveillance video frame image.

[0037] Step S150, binarize the PSNR result to obtain a binarized image corresponding to the PSNR result matrix, that is, a binarized image corresponding to the difference image obtained according to the monitoring video frame image; wherein the PSNR result is binarized by setting a PSNR threshold parameter, such as Figure 5 As shown, Figure 4 The PSNR result shown is a PSNR result matrix after binarization using a PSNR threshold. In this embodiment, a first initial preset value is pre-set as the initial value of the PSNR threshold, and then the PSNR threshold is automatically adjusted using a first preset rule, wherein the first preset rule is that the first initial preset value increases or decreases step by step according to the first preset value, and a maximum and a minimum value are preset, and the PSNR threshold value ranges from 12 to 36; in addition, since the background color in the frame image of the original surveillance video is not uniform, only using a single PSNR threshold cannot ensure that the tracking target can be successfully obtained. Therefore, in the process of binarizing the PSNR result, a group of PSNR thresholds are actually used. When binarizing the PSNR result, multiple levels of PSNR thresholds are used to generate multiple binarized images with different confidence levels, so as to avoid the adverse effects that fixed parameters may cause under different lighting conditions.

[0038] Step S200, obtaining a plurality of rectangular blocks based on the bright blocks in the binary image, effectively improving the calculation efficiency of the method of the present invention; performing horizontal and vertical summation on the binary image, and obtaining a plurality of rectangular blocks based on the bright blocks in the binary image, wherein the bright blocks correspond to image blocks in the binary image whose binarization results are output as 1.

[0039] In one embodiment, the method of obtaining a plurality of rectangular blocks based on the bright blocks in the binary image in step S200 specifically includes the following steps:

[0040] Step S210, summing up the binarized image in columns, discarding the columns whose summation results are 0, and combining the adjacent columns whose summation results are not 0 to obtain a plurality of long rectangular blocks;

[0041] Step S220, summing up each long rectangular block in rows, discarding rows with summation results of 0, to obtain a number of quasi-rectangular blocks; wherein the quasi-rectangular block is a rectangular block composed of a plurality of bright blocks and non-bright blocks or a plurality of bright blocks;

[0042] Step S230, summing up each quasi-rectangular block in columns, discarding columns with summation results of 0, and obtaining a plurality of rectangular blocks; wherein the rectangular block is composed of at least one bright block.

[0043] Step S300, establish a set of objects to be selected based on the rectangular blocks in the current frame binary image, perform matching calculations on the objects to be selected and the nodes in the matching pool one by one, generate new trajectory tree nodes according to the calculation results, and then update the matching pool and the matching forest; wherein, the references of the trajectory tree nodes generated in the previous P frame binary images of the current frame binary image have entered the matching pool, and when processing the objects to be selected of the current binary image, the nodes that have been successfully matched in the previous P-1 rounds of matching calculations have left the matching pool, and P is a first preset threshold; nodes are formed corresponding to the objects to be selected in the binary image, and the nodes in the matching pool and the nodes on the trajectory tree all use the nodes formed corresponding to the objects to be selected by reference or pointer, In this embodiment and the following description, the reference of the trajectory tree node is simplified to the trajectory tree node, and the reference of the trajectory tree node in the matching pool is simplified to the node in the matching pool; the number of nodes in the matching pool is initially empty, the matching forest is composed of trajectory trees, the number of trajectory trees in the matching forest is initially empty, the trajectory tree is composed of trajectory tree nodes connected in a tree-like manner, and at least one trajectory tree node is required as the root node to establish a trajectory tree; specifically, the matching pool is maintained when the binary image matching calculation of each frame is completed. For example, when calculating the n+1th frame, the matching pool maintained after the binary image matching calculation of the nth frame is completed is used, so the current matching pool does not contain the reference to the trajectory tree node generated by the current frame binary image; each After the frame binary image is processed, the trajectory tree and the matching forest will be updated. The nodes in the matching pool and the nodes of the trajectory tree are transformed from the candidate objects. In this embodiment, the nodes in the matching pool represent the targets that may belong to the high-altitude parabolic tracking; the new nodes of the matching forest may be new child nodes of the existing trajectory tree, or they may be the root nodes of the newly created trajectory tree. The trajectory tree is used instead of the linear table to associate the candidate object results, and the several tree branches below the trajectory tree can be matched and associated with the rectangular blocks calculated from the new frame binary image, so as to generate new child nodes of the trajectory tree. The purpose is to retain multiple possibilities of target movement in the continuous image processing process and postpone judgment. The trajectory tree grows with the changes of the binary image until Confirm that the changes of all tree branches of the trajectory tree are completed, and then evaluate all tree branches of the trajectory tree to find the most reliable parabolic path from the root node of the trajectory tree to all tree branches, thereby avoiding erroneous results due to local or temporary interference in the intermediate process; in addition, the frame number of the current frame binary image corresponds to the frame number of the difference image, and the rectangular blocks in the current frame binary image obtained in step S200 correspond one-to-one to the objects to be selected in the current frame binary image. The matching pool cooperates with the growth of the trajectory tree through a linear table, thereby improving the efficiency of the trajectory tree matching growth, avoiding frequent traversal of all trajectory trees in the matching forest, and effectively reducing the number of traversals, and improving the matching efficiency of the objects to be selected in the current frame binary image.

[0044] Specifically, in the initial state, the number of nodes in the matching pool is 0. When there are A1 objects to be selected in the first frame of the binary image and they are associated with the nodes in the matching pool, since there are no nodes in the initial matching pool, after the association is completed, new nodes corresponding to the A1 objects to be selected will be added to the matching pool, and the nodes corresponding to the A1 objects to be selected will correspond to the root nodes of the A1 newly created trajectory trees; when there are A2 objects to be selected in the second frame of the binary image and they are associated with the nodes in the matching pool, when a2 of the A2 objects to be selected are successfully associated with the nodes in the matching pool, after the association is completed, at least A2 will be added to the matching pool. A new node, a2 successfully associated candidates will generate corresponding child nodes on the existing trajectory tree. Specifically, at least A2 nodes will be added to the matching pool after the association is completed. The associated candidates will generate nodes equivalent to the number of successful associations, and these successfully generated nodes will be connected to the corresponding parent nodes. These successfully generated nodes enter the matching pool and move their parent nodes out of the matching pool; the unassociated candidates generate a corresponding node and enter the matching pool; and so on, the same applies to the candidates in the third frame, the fourth frame, ..., and the Nth frame binarized images when they are associated with the nodes in the matching pool.

[0045] Among them, the attribute information of the object to be selected includes the frame sequence number information of the current frame binary image in the processed video stream, the specific position information of the rectangular block in the current frame binary image, the width and height information of the rectangular block, etc. When overlapping objects to be selected are generated due to the use of multiple PSNR thresholds, the attribute information of the object to be selected corresponding to the minimum PSNR threshold among the overlapping objects to be selected is selected; specifically, a candidate object is established for each output rectangular block, and its attribute information includes the frame sequence number of the current frame binary image in the processed video stream, the rectangular block position, the width and height information, etc., and the new rectangular block does not pursue a one-to-one result similar to the Hungarian algorithm during the implementation of the detection and tracking method of the present invention, but generates a child node for each possible parent node, avoiding the situation of local optimality but global failure, and achieving the effect of connecting as many moving nodes as possible, and its purpose is to make the motion trajectory of the object as complete as possible, thereby avoiding the occurrence of false alarms such as the presence of a section that meets the characteristics of a parabolic object in the complex motion trajectory of a flying bird or insect.

[0046] The method of step S300 specifically includes the following steps:

[0047] Step S310, establishing a to-be-selected object based on the rectangular block;

[0048] Step S320, matching the candidate object A in the current frame binary image with the nodes in the matching pool one by one, and determining whether there is a node B in the matching pool that matches the candidate object A in the current frame binary image; if so, proceeding to step S330, if not, proceeding to step S340;

[0049] Step S330: If there is a node B in the matching pool that is matched with the candidate object A in the current frame binary image, a node corresponding to the candidate object A is generated on the existing trajectory tree as a child node of node B, and after the matching attempts of all the candidate objects in the current frame binary image are completed, the node corresponding to the candidate object A is added to the matching pool;

[0050] Step S340, if there is no node B in the matching pool that is matched and associated with the candidate object A in the current frame binary image, then after all the candidate objects in the current frame binary image are matched and associated, a corresponding trajectory tree node is newly created according to the candidate object A as the root node of the newly created trajectory tree, and added to the matching pool, and the process proceeds to step S350; wherein, the operation of adding the node corresponding to the candidate object A in the current frame binary image to the matching pool is performed after all the candidate objects in the current frame binary image are matched and associated with the nodes in the matching pool, thereby avoiding unnecessary redundant calculation caused by the node corresponding to the candidate object A in the current frame binary image entering the matching pool.

[0051] Step S350: if, since the node in the matching pool is generated, in the process of calculating and processing the subsequent P-frame binary images corresponding to the node, no candidate object that successfully matches the node is found, then the node is no longer associated with the matching and is removed from the matching pool; if, since the node in the matching pool is generated, the node is first successfully matched with the candidate object of one of the subsequent P-frame binary images, then after the matching calculation of all nodes in the matching pool with the candidate object of the one of the binary images is completed, the node is removed from the matching pool, thereby realizing the update operation of the matching forest; since the node in the matching pool is generated, it is necessary to calculate and process the subsequent P-frame binary images of the frame binary image corresponding to the node, and the node is processed according to the actual matching result during the calculation process; wherein, P is a first preset threshold value, and the completion of the matching calculation refers to that after the associated matching calculation of all the candidate objects in the frame binary image and all the nodes in the current matching pool is performed, in this embodiment, the maximum number of frames that the node can continuously match in the matching pool is controlled by the first preset threshold value P number of frames.

[0052] In this embodiment, the node data of the trajectory tree can record the attribute information of the node, such as the position, width, height, speed, acceleration, motion angle, angular velocity, etc., and all the child nodes matched with the node. The calculation results of the attributes such as the speed, acceleration, motion angle, angular velocity, etc. of the node of the trajectory tree can be obtained by the calculation method of the speed, acceleration, motion angle, and angular velocity values ​​of the tracking target in the x / y direction when the selected object A and the node B belong to the same tracking target in step S3301; when the trajectory tree is a newly created trajectory tree, since the newly created trajectory tree lacks the pre-information provided by the parent node, the initial expected range parameter of the frame binarization image corresponding to the root node of the newly created trajectory tree can be set by a circle with a preset value as the radius, the first-level child node of the root node of the trajectory tree uses the average speed instead of the speed, does not calculate the acceleration, and calculates all the parameters such as the speed, acceleration, motion angle, and angular velocity values ​​from the second-level child node; in addition, when the matching calculation of each frame of the binary image is completed and the expected range of the node is updated, if the root node is not matched, the radius of the expected range will be increased.

[0053] Furthermore, in the present embodiment, there may be a situation in the matching pool where a node B can match and be associated with multiple candidate objects in the current frame binary image. At this time, multiple nodes corresponding to the candidate objects A1 and A2 in the current frame binary image are generated on the existing trajectory tree as child nodes of node B, that is, when the node of the existing trajectory tree is node B, multiple branches with node B as the parent node will be formed, and the nodes corresponding to the candidate objects A1 and A2 in the current frame binary image will become child nodes of node B respectively; there may also be a situation in the matching pool where multiple nodes B1 and B2 can match and be associated with a candidate object A in the current frame binary image. At this time, multiple nodes with the nodes corresponding to the candidate object A as child nodes are generated on the existing trajectory tree respectively; specifically, when multiple nodes B1 and B2 are parent nodes respectively, multiple nodes with the nodes corresponding to the candidate object A as child nodes are generated on the existing trajectory tree, such as Figure 6 As shown, if all nodes in the lower three layers of sequence frames in one trajectory tree have corresponding matching associated candidate objects, a many-to-many relationship can be formed between the candidate objects and the nodes in the matching pool.

[0054] In one embodiment, in step S330, if there is a node B in the matching pool that matches and is associated with the candidate object A in the binary image of the current frame, a node corresponding to the candidate object A is generated on the existing trajectory tree as a child node of the node B, and the specific operations are:

[0055] Step S3301, determine whether the candidate object A in the current frame binary image is within the expected range of the frame binary image corresponding to the node B of the trajectory tree; if so, obtain the speed, acceleration, motion angle, and angular velocity values ​​of the tracking target in the x / y direction when the candidate object A and the node B belong to the same tracking target according to the displacement and frame sequence number difference between the candidate object A and the node B. When the speed, acceleration, motion angle, and angular velocity values ​​of the tracking target meet the preset conditions, the candidate object A is matched and associated with the node B, and a node corresponding to the candidate object A is generated on the trajectory tree as a child node of the node B, that is, the node corresponding to the candidate object A is matched and associated with the node B. The node is also used as a new node in the matching forest, and the calculated data such as speed, acceleration, motion angle, angular velocity value, etc. are saved as node attribute information to avoid repeated calculation, thereby improving the calculation efficiency of the method of the present invention; wherein, the trajectory tree associated with the node B in the matching pool may be a newly created trajectory tree with the node B as the root node, or it may be an existing trajectory tree with the node B as the child node. In fact, as long as it is ensured that the time the node stays in the matching pool is always less than the time it exists on the trajectory tree, the nodes in the matching pool and the nodes on the trajectory tree can all use the nodes corresponding to the objects to be selected in the binary image by reference or pointer.

[0056] Among them, in the step S3301, according to the displacement and frame number difference between the candidate object A and the node B, the speed, acceleration, motion angle, and angular velocity values ​​of the tracking target in the x / y direction are obtained when the candidate object A and the node B belong to the same tracking target. When the speed, acceleration, motion angle, and angular velocity values ​​of the tracking target all meet the preset conditions, the candidate object A is matched and associated with the node B, and a node corresponding to the candidate object A is generated on the trajectory tree as a child node of the node B. The method includes the following steps:

[0057] Assume that the frame numbers of the binary images corresponding to the selected object A and node B are n and m respectively, then:

[0058] D x，n，m =X n -X m (1)

[0059] D y，n，m =Y n -Y m (2)

[0060] Formula (1) and formula (2) can be used to calculate the pixel distance that the tracking target moves on the image between the mth frame and the nth frame, (X m , Y m ) and (X n , Y n) represent the horizontal and vertical coordinates of the tracking target in the mth frame and the nth frame respectively, based on the coordinates of the center point of the rectangular block, t n，m is the time interval between the mth frame and the nth frame; here, the pixel distance that the tracking target moves on the image between the mth frame and the nth frame is not used to calculate the final target speed, because this data can only reflect the average speed of the tracking target during the time period from the mth frame to the nth frame, and when the mth frame and the nth frame are not adjacent, this data is used to calculate the final target speed of the tracking target, which will lead to error amplification.

[0061] Assume that the frame numbers of the binary images corresponding to the selected object A and node B are n and m respectively, then:

[0062]

[0063]

[0064] Among them, t n，m is the time interval between the mth frame and the nth frame, (V x，m , V y，m ) is the lateral and longitudinal speed of the tracking target in the mth frame. Formula (3) and formula (4) are used to calculate the acceleration (a) of the tracking target in the lateral and longitudinal directions during the period from the mth frame to the nth frame. x，n，m , a y，n，m ).

[0065] In addition, the lateral acceleration is not directly assigned to 0 because, in addition to accidental factors such as wind and collision, in engineering practice, due to objective conditions, the installation angle of the surveillance camera often causes the frame binary image of the surveillance video shot in the direction of gravity to be not vertical. Through such objective calculation, the impact on the judgment result of the tracking target can be reduced.

[0066] Assume that the frame numbers of the binary images corresponding to the selected object A and node B are n and m respectively, then:

[0067] V x，n =V x，m +a x，n，m t n，m (5)

[0068] V y，n =V y，m +a y，n，m t n，m (6)

[0069] Based on the acceleration (a x，n，m , a y，n，m ), the speed of the mth frame (V x，m , V y，m) and the time interval t from the mth frame to the nth frame n，m , the lateral and longitudinal velocities (V x，n , V y，n ).

[0070] Assume that the frame numbers of the binary images corresponding to the selected object A and node B are n and m respectively, then:

[0071] θ n =arctan(V y，n , V x，n ) (7)

[0072] The motion angle θ of the target in the nth frame is calculated by formula (7): n , and then according to the position relationship between the selected object A and node B in the coordinate system, θ n Convert to the interval [0,2π).

[0073] Assume that the frame numbers of the binary images corresponding to the selected object A and node B are n and m respectively, then:

[0074] ω n，m =(θ m -θ m ) / t n，m (8)

[0075] The angular velocity ω of the tracked target between the mth frame and the nth frame is calculated by formula (8): n，m ; In engineering practice, it is assumed that the rotation angle of the tracked target will not exceed 180 degrees during the period from the mth frame to the nth frame.

[0076] After obtaining the speed, acceleration, motion angle, angular velocity and other information of the tracked target in the mth frame and the nth frame through formula (1) to formula (8), it is determined whether the speed, acceleration, deflection angle and angular velocity of the tracked target between the mth frame and the nth frame are within the preset speed threshold, preset acceleration threshold, preset deflection angle threshold and preset angular velocity threshold. If so, it is determined that the tracked target in the nth frame is the successor of the tracked target in the mth frame, and the candidate object A is matched and associated with the node B. A node corresponding to the candidate object A is generated on the trajectory tree as a child node of the node B, and also as a new node in the matching forest. The deflection angle can be calculated by the difference in the motion angle of the tracked target between the mth frame and the nth frame.

[0077] In one embodiment, the method for determining whether the candidate object A in the current frame binary image is within the expected range of the frame binary image where the node B of the trajectory tree is located in step S3301 is specifically performed as follows:

[0078] The distance from the center of the candidate object A to the expected center of the node B is calculated, and then the distance is compared with the preset distance. If the distance is greater than the preset distance, the candidate object A in the current frame binary image is not within the expected range of the frame binary image corresponding to the node B; if the distance is not greater than the preset distance, the candidate object A in the current frame binary image is within the expected range of the frame binary image corresponding to the node B; wherein the distance from the center of the candidate object A to the expected center of the node B can be obtained by (X m , Y m ) and (X n , Y n ) represent the horizontal and vertical coordinates of the tracking target on the mth frame and the nth frame respectively.

[0079] Step S400, judging whether the trajectory tree meets the preset conditions for deletion. If the trajectory tree meets the preset conditions for deletion, the trajectory tree is determined to be a trajectory tree to be deleted; wherein, only after it is determined that a trajectory tree has finished growing, the number of nodes passed on the path between the root node of the trajectory tree and the leaf node of the trajectory tree is calculated, and the trajectory path that meets the parabolic characteristics in the trajectory tree to be deleted enters the next candidate operation.

[0080] The method of determining whether the trajectory tree satisfies the preset to-be-deleted condition in step S400, and determining the trajectory tree as a trajectory tree to be deleted if the trajectory tree satisfies the preset to-be-deleted condition, specifically comprises the following operations:

[0081] The attribute information of the track tree node contains two timestamps, i.e., frame number information. One is the creation timestamp, which is assigned when the node is created and will not change afterwards. The other is the update timestamp. Whenever a track tree node is connected to a new child node, the timestamp of the node and all the direct ancestor nodes of the node are all updated to the creation timestamp of the new child node. Among them, the update timestamp of the newly created node is equal to the creation timestamp.

[0082] When traversing the matching forest, each trajectory tree is traversed first and foremost. When a timestamp difference of one of the nodes of the trajectory tree is greater than Q, and the node is the root node of the trajectory tree, the trajectory tree corresponding to the node is determined to be the trajectory tree to be deleted, wherein the timestamp difference = current timestamp - update timestamp, the current timestamp refers to the frame sequence information corresponding to the current frame binary image when traversing the matching forest, Q is the second preset threshold, Q>P; in fact, after the matching calculation of the object to be selected in the current frame binary image is completed, the matching forest is traversed. As long as it is ensured that the time the node stays in the matching pool is always less than the time it exists on the trajectory tree, that is, Q>P, the nodes in the matching pool and the nodes on the trajectory tree can all use the nodes corresponding to the object to be selected in the binary image by reference or pointer.

[0083] In one embodiment, the step S400 further includes

[0084] It is judged whether the tree branch of the trajectory tree meets the pruning condition. If the pruning condition is met, the tree branch is deleted from the trajectory tree to implement the pruning operation of each trajectory tree in the matching forest. Specifically, after completing the matching association operation between the candidate object of the binary image and the node in the matching pool, the trajectory trees in the matching forest are traversed to implement the pruning operation of each trajectory tree in the matching forest, and the tree branches that are no longer likely to belong to the parabolic path are excluded, thereby avoiding the infinite expansion and association of the trajectory trees in the matching forest, thereby effectively reducing the calculation pressure.

[0085] Furthermore, in step S400, it is determined whether the tree branch of the trajectory tree meets the pruning condition. If the pruning condition is met, the method of deleting the tree branch on the trajectory tree specifically comprises the following operations:

[0086] The attribute information of the trajectory tree node contains two timestamps, i.e., frame number information. One is the creation timestamp, which is assigned when the node is created and will not change afterwards. The other is the update timestamp. Whenever a trajectory tree node accesses a new child node, the update timestamps of the node itself and all the direct ancestor nodes of the node are updated to the creation timestamp of the new child node. Among them, the update timestamp of the newly created node is equal to the creation timestamp.

[0087] When traversing the matching forest, each trajectory tree is traversed first and foremost. When the timestamp difference of one of the nodes of the trajectory tree is greater than Q, and the one of the nodes is not the root node of the trajectory tree, the tree branch corresponding to the one of the nodes is pruned, wherein the timestamp difference = current timestamp - update timestamp, the current timestamp refers to the frame sequence information corresponding to the current frame binary image when traversing the matching forest, Q is the second preset threshold, Q>P, and the pruning operation of the trajectory tree in the matching forest is realized; the tree branch corresponding to the one of the nodes includes the one of the nodes and all its direct descendant nodes, but does not include its father, grandfather, brothers and brothers' descendant nodes; after the matching calculation of the object to be selected in the current frame binary image is completed, the traversal operation of the matching forest is started.

[0088] In this embodiment, judging whether a tree branch of a trajectory tree satisfies a pruning condition and judging whether a trajectory tree satisfies a preset condition to be deleted are implemented simultaneously. Specifically, the attribute information of a trajectory tree node includes two timestamps, namely, frame number information. One is a creation timestamp, namely, which frame of binary image is generated by the node, which is assigned when the node is created and will not change afterwards. The second is an update timestamp (the frame number corresponding to the last update of the tree branch). Whenever a trajectory tree node accesses a new child node, the update timestamp of the node itself and all the direct ancestor nodes of the node are all updated to the creation timestamp of the new child node. Among them, the update timestamp of the newly created node is equal to the creation timestamp. When traversing the matching forest, each trajectory tree is traversed first and foremost. When encountering a node with a timestamp difference value greater than Q, if the node is not a root node, it is judged that the tree branch corresponding to the node needs to be pruned. If the node is a root node, it is judged that the trajectory tree corresponding to the node needs to be deleted.

[0089] Step S500, obtaining all trajectory paths corresponding to the trajectory tree to be deleted, judging whether there is a trajectory path in the trajectory path corresponding to the trajectory tree to be deleted that satisfies the preset suspected parabolic condition, if there is a trajectory path that satisfies the preset suspected parabolic condition, then judging the trajectory path as a suspected parabolic path, and saving the suspected parabolic path to obtain a saved suspected parabolic path set, and then removing all trajectory trees to be deleted from the matching forest; wherein, when judging all trajectory paths corresponding to the trajectory tree to be deleted, it is necessary to calculate the trajectory path composed of the marked nodes on each tree branch from the root node of the trajectory tree to be deleted to the trajectory tree to be deleted, wherein the marked node is the node corresponding to the timestamp difference (update timestamp-creation timestamp) of the node encountered during the traversal process that is less than Z, so as to avoid occasional interference at the end of the trajectory path and affect the judgment result, Z is the third preset threshold, generally taking a value of 1 to 4, preferably 3.

[0090] In one embodiment, in step S500, it is determined whether there is a trajectory path in the trajectory path corresponding to the trajectory tree to be deleted that satisfies a preset suspected parabolic condition, and if there is a trajectory path that satisfies the preset suspected parabolic condition, the method for determining that the trajectory path is a suspected parabolic path specifically comprises the following operations:

[0091] The trajectory path whose total number of nodes is higher than the preset node number threshold and whose acceleration of nodes along the way is lower than the preset acceleration threshold, whose angular velocity of nodes along the way is lower than the preset angular velocity threshold, whose deflection angle of nodes along the way is lower than the preset deflection angle threshold is selected as the suspected parabolic path; wherein, the total number of nodes on the trajectory path refers to the number of nodes passed through on the path from the root node to the marked node of the trajectory tree corresponding to the trajectory path, and the marked node is the node corresponding to the timestamp difference (update timestamp-creation timestamp) of the node encountered during the traversal process is less than Z, so as to avoid the influence caused by the short passage of moths, birds, etc. on the trajectory path; specifically, the method of obtaining the trajectory path during the traversal process is to build a linear list as a stack, and then adopt the root-first traversal method, and whenever a node is visited, it is pushed into the stack, and the nodes in the stack are arranged in order from the bottom of the stack to the top of the stack, which is the path from the root node to the current node; then all its subtrees are traversed in turn; finally, it is popped out of the stack and turned to visit its sibling nodes or return to the parent node. In addition, the speed, acceleration, deflection angle, and angular velocity values ​​corresponding to the nodes on the trajectory path can be obtained by the calculation method of the speed, acceleration, motion angle, and angular velocity values ​​of the tracking target in the x / y direction when the selected object A and node B belong to the same tracking target in step S3301.

[0092] like Figure 2 As shown, in one embodiment, after step S500, the method further includes:

[0093] Step S500-1, when the traversal process visits one of the nodes on the trajectory tree, if the node is still in the matching pool, the expected range parameters of the binary image frame corresponding to the node are updated to facilitate matching and associating with the candidate object in the next binary image frame, thereby obtaining a new node of the matching forest, and then proceeding to execute step S300, so as to avoid the candidate object from colliding or being disturbed by wind and being unable to match and associate with the node of the trajectory tree; wherein, the traversal process is to perform a pre-root traversal on each trajectory tree when traversing the matching forest.

[0094] In this embodiment, the step S500 - 1 is located after the step S500 , and the step S500 - 1 can also be described as being located before the step S300 , and the step S500 - 1 can also be implemented synchronously with the step S300 .

[0095] Step S600: determine whether the stored suspected parabolic path meets the preset parabolic condition. If there is a suspected parabolic path that meets the preset parabolic condition, determine that the suspected parabolic path is a parabolic event.

[0096] In one embodiment, in step S600, it is determined whether the stored suspected parabolic path satisfies the preset parabolic condition. If there is a suspected parabolic path that satisfies the preset parabolic condition, the method for determining that the suspected parabolic path is a parabolic event specifically comprises the following operations:

[0097] Among the saved suspected parabolic paths, the trajectory path with a total number of nodes higher than a preset node number threshold and the smallest acceleration statistical variance of the nodes along the way is selected as a parabolic event; wherein, the total number of nodes on the trajectory path refers to the number of nodes passed on the path from the root node to the marked node of the trajectory tree corresponding to the trajectory path, and the marked node is the node corresponding to the timestamp difference (update timestamp-creation timestamp) of the node encountered during the traversal process that is less than Z; the speed, acceleration, motion angle, and angular velocity values ​​corresponding to the nodes on the trajectory path can be obtained by the calculation method of the speed, acceleration, motion angle, and angular velocity values ​​of the tracking target in the x / y direction when the selected object A and node B belong to the same tracking target in step S3301.

[0098] like Figure 3 As shown, in order to make the technical solution of the present invention clearer, a preferred embodiment is described below.

[0099] Step S100, obtaining a binary image corresponding to a surveillance video frame image;

[0100] Step S200, obtaining a plurality of rectangular blocks based on bright blocks in the binary image;

[0101] Step S300, establishing a set of objects to be selected based on the rectangular blocks in the current frame binary image, performing matching calculations on the objects to be selected and the nodes in the matching pool one by one, generating new trajectory tree nodes according to the calculation results, and then updating the matching pool and the matching forest; wherein the number of nodes in the matching pool is initially empty, the matching forest is composed of trajectory trees, the number of trajectory trees in the matching forest is initially empty, and the trajectory tree is composed of trajectory tree nodes connected in a tree-like manner;

[0102] Step S400, judging whether the trajectory tree meets the preset to-be-deleted condition, if the trajectory tree meets the preset to-be-deleted condition, then judging the trajectory tree as the trajectory tree to be deleted; judging whether the tree branch of the trajectory tree meets the pruning condition, if the pruning condition is met, then deleting the tree branch on the trajectory tree;

[0103] Step S500, obtaining all trajectory paths corresponding to the trajectory tree to be deleted, determining whether there is a trajectory path in the trajectory path corresponding to the trajectory tree to be deleted that satisfies a preset suspected parabolic condition, and if there is a trajectory path that satisfies the preset suspected parabolic condition, determining the trajectory path to be a suspected parabolic path, and saving the suspected parabolic path to obtain a saved suspected parabolic path set, and then removing all trajectory trees to be deleted from the matching forest;

[0104] Step S500-1: when the traversal process accesses one of the nodes on the trajectory tree, if the node is still in the matching pool, update the expected range parameter of the binary image of the frame corresponding to the node;

[0105] Step S600: determine whether the stored suspected parabolic path meets the preset parabolic condition. If there is a suspected parabolic path that meets the preset parabolic condition, determine that the suspected parabolic path is a parabolic event.

[0106] In summary, a high-altitude parabolic detection and tracking method of the present invention uses a trajectory tree to determine and filter the suspected parabolic path of the tracking target, thereby finding out the confirmed parabolic event, effectively reducing the probability of missed reports and false alarms, and controlling the influence of high-frequency noise in the frame image by dividing the differential image into several image blocks.

[0107] Although the embodiments of the present invention have been described in conjunction with the accompanying drawings, those skilled in the art may make various modifications and variations without departing from the spirit and scope of the present invention, and such modifications and variations are all within the scope defined by the appended claims.

Claims

1. A method for detecting and tracking objects thrown from a high altitude, characterized in that: The steps include: Step S100, obtaining a binary image corresponding to a surveillance video frame image; Step S200, obtaining a plurality of rectangular blocks based on bright blocks in the binary image; Step S300, establishing a set of objects to be selected based on the rectangular blocks in the current frame binary image, performing matching calculations on the objects to be selected and the nodes in the matching pool one by one, generating new trajectory tree nodes according to the calculation results, and then updating the matching pool and the matching forest; wherein the number of nodes in the matching pool is initially empty, the matching forest is composed of trajectory trees, the number of trajectory trees in the matching forest is initially empty, and the trajectory tree is composed of trajectory tree nodes connected in a tree-like manner; Step S310, establishing a to-be-selected object based on the rectangular block; Step S320, matching the candidate object A in the current frame binary image with the nodes in the matching pool one by one, and determining whether there is a node B in the matching pool that matches the candidate object A in the current frame binary image; if so, proceeding to step S330, if not, proceeding to step S340; Step S330: If there is a node B in the matching pool that is matched with the candidate object A in the current frame binary image, a node corresponding to the candidate object A is generated on the existing trajectory tree as a child node of node B, and after the matching attempts of all the candidate objects in the current frame binary image are completed, the node corresponding to the candidate object A is added to the matching pool; Step S340: If there is no node B in the matching pool that matches the candidate object A in the current frame binary image, after all the candidate objects in the current frame binary image are matched and associated, a corresponding trajectory tree node is newly created according to the candidate object A as the root node of the newly created trajectory tree, and added to the matching pool, and the process proceeds to step S350; Step S350: if, since the node in the matching pool is generated, in the process of calculating and processing the subsequent P-frame binary images corresponding to the node, no candidate object that successfully matches the node is found, then no association matching is performed on the node, and the node is removed from the matching pool; if, since the node in the matching pool is generated, the node is first successfully matched with the candidate object of one of the subsequent P-frame binary images, then after the matching calculation of all nodes in the matching pool and the candidate object of one of the binary images is completed, the node is removed from the matching pool; wherein P is a first preset threshold; Step S400, determining whether the trajectory tree meets the preset to-be-deleted condition, if the trajectory tree meets the preset to-be-deleted condition, determining the trajectory tree as a trajectory tree to be deleted; The attribute information of the track tree node contains two timestamp information. One is the creation timestamp, which is assigned when the node is created and will not change afterwards. The other is the update timestamp. Whenever a track tree node accesses a new child node, the update timestamp of the node itself and all the direct ancestor nodes of the node are updated to the creation timestamp of the new child node. Among them, the update timestamp of the newly created node is equal to the creation timestamp. When traversing the matching forest, perform a pre-root traversal on each trajectory tree. When a timestamp difference of one of the nodes of the trajectory tree is greater than Q, and the one of the nodes is the root node of the trajectory tree, the trajectory tree corresponding to the one of the nodes is determined as the trajectory tree to be deleted, where the timestamp difference = current timestamp - update timestamp, the current timestamp refers to the frame sequence information corresponding to the current frame binarization image when traversing the matching forest, and Q is the second preset threshold; Step S500, obtaining all trajectory paths corresponding to the trajectory tree to be deleted, determining whether there is a trajectory path in the trajectory path corresponding to the trajectory tree to be deleted that satisfies a preset suspected parabolic condition, and if there is a trajectory path that satisfies the preset suspected parabolic condition, determining the trajectory path to be a suspected parabolic path, and saving the suspected parabolic path to obtain a saved suspected parabolic path set, and then removing all trajectory trees to be deleted from the matching forest; Step S600, determine whether the saved suspected parabolic path meets the preset parabolic conditions. If there is a suspected parabolic path that meets the preset parabolic conditions, determine that the suspected parabolic path is a parabolic event: among the saved suspected parabolic paths, select the trajectory path with a total number of nodes higher than the preset node number threshold and the smallest acceleration statistical variance of the nodes along the way as a parabolic event.

2. The method for detecting and tracking objects thrown from a high altitude according to claim 1, characterized in that: The method of step S100 comprises the following steps: Step S110, obtaining a frame image of a surveillance video; Step S120, obtaining a differential image corresponding to the surveillance video frame image; Step S130, dividing the difference image into several image blocks, and obtaining the PSNR result corresponding to each image block; Step S140, arranging the PSNR results corresponding to each image block in a matrix manner to form a PSNR result matrix; Step S150: binarize the PSNR result matrix to obtain a binarized image corresponding to the PSNR result matrix.

3. The method for detecting and tracking high-altitude objects according to claim 1 or 2, characterized in that: The method of step S200 comprises the following steps: Step S210, summing up the binarized image in columns, discarding the columns whose summation results are 0, and combining the adjacent columns whose summation results are not 0 to obtain a plurality of long rectangular blocks; Step S220, summing up each long rectangular block in rows, discarding rows with summation results of 0, to obtain a number of quasi-rectangular blocks; wherein the quasi-rectangular block is a rectangular block composed of a plurality of bright blocks and non-bright blocks or a plurality of bright blocks; Step S230, summing up each quasi-rectangular block in columns, discarding columns with summation results of 0, and obtaining a plurality of rectangular blocks; wherein the rectangular block is composed of at least one bright block.

4. The method for detecting and tracking objects thrown from a high altitude according to claim 1, characterized in that: In step S330, if there is a node B in the matching pool that matches and is associated with the candidate object A in the binary image of the current frame, a method of generating a node corresponding to the candidate object A on the existing trajectory tree as a child node of node B is specifically performed as follows: Step S3301, determining whether the object A to be selected in the current frame binary image is within the expected range of the frame binary image corresponding to the node B of the trajectory tree; If so, based on the displacement and frame number difference between the candidate object A and node B, the speed, acceleration, motion angle, and angular velocity values ​​of the tracking target in the x / y direction are obtained when the candidate object A and node B belong to the same tracking target. When the speed, acceleration, motion angle, and angular velocity values ​​of the tracking target meet the preset conditions, the candidate object A is matched and associated with node B, and a node corresponding to the candidate object A is generated on the trajectory tree as a child node of node B.

5. The method for detecting and tracking objects thrown from a high altitude according to claim 4, characterized in that: In the step S3301, according to the displacement and frame number difference between the candidate object A and the node B, the speed, acceleration, motion angle, and angular velocity values ​​of the tracking target in the x / y direction are obtained when the candidate object A and the node B belong to the same tracking target. When the speed, acceleration, motion angle, and angular velocity values ​​of the tracking target all meet the preset conditions, the candidate object A is matched and associated with the node B, and a node corresponding to the candidate object A is generated on the trajectory tree as a child node of the node B, comprising the following steps: Assume that the frame numbers of the binary images corresponding to the selected object A and node B are n and m respectively, then: D x，n，m =X n -X m (1) D y，n，m =Y n -Y m (2) The pixel distance that the tracking target moves on the image between the mth frame and the nth frame is calculated by formula (1) and formula (2), (X m , Y m ) and (X n , Y n ) represent the horizontal and vertical coordinates of the tracking target in the mth frame and the nth frame respectively, based on the coordinates of the center point of the rectangular block, t n，m is the time interval between the mth frame and the nth frame; Assume that the frame numbers of the binary images corresponding to the selected object A and node B are n and m respectively, then: Among them, t n，m is the time interval between the mth frame and the nth frame, (V x，m , V y，m ) is the lateral and longitudinal speed of the tracking target in the mth frame. Formula (3) and formula (4) are used to calculate the acceleration (a) of the tracking target in the lateral and longitudinal directions during the period from the mth frame to the nth frame. x，n，m , a y，n，m ); Assume that the frame numbers of the binary images corresponding to the selected object A and node B are n and m respectively, then: V x，n =V x，m +a x，n，m t n，m (5) V y，n =V y，m +a y，n，m t n，m (6) Based on the acceleration (a x，n，m , a y，n，m ), the speed of the mth frame (V x，m , V y，m ) and the time interval t from the mth frame to the nth frame n，m , the lateral and longitudinal velocities (V x，n , V y，n ); Assume that the frame numbers of the binary images corresponding to the selected object A and node B are n and m respectively, then: θ n =arctan(V y，n ,V x，n ) (7) The motion angle θ of the target in the nth frame is calculated by formula (7): n , and then according to the position relationship between the selected object A and node B in the coordinate system, θ n Convert to the interval [0, 2π); Assume that the frame numbers of the binary images corresponding to the selected object A and node B are n and m respectively, then: oh n，m =(θ n -θ m ) / t n，m (8) The angular velocity ω of the tracked target between the mth frame and the nth frame is calculated by formula (8): n，m ; After obtaining the speed, acceleration, motion angle, and angular velocity numerical information of the tracked target in the mth frame and the nth frame through formula (1) to formula (8), it is determined whether the speed, acceleration, deflection angle, and angular velocity of the tracked target between the mth frame and the nth frame are within the range of the preset speed threshold, the preset acceleration threshold, the preset deflection angle threshold, and the preset angular velocity threshold. If so, it is determined that the tracked target in the nth frame is the successor of the tracked target in the mth frame, and the candidate object A is matched and associated with the node B, and a node corresponding to the candidate object A is generated on the trajectory tree as a child node of the node B.

6. The method for detecting and tracking high-altitude objects according to claim 1 or 2, characterized in that: The step S400 also includes Determine whether the tree branch of the trajectory tree meets the pruning condition. If the pruning condition is met, delete the tree branch on the trajectory tree. The method of determining whether a tree branch of the trajectory tree satisfies a pruning condition in step S400, and if the pruning condition is satisfied, deleting the tree branch on the trajectory tree, specifically comprises the following operations: The attribute information of the track tree node contains two timestamp information. One is the creation timestamp, which is assigned when the node is created and will not change afterwards. The other is the update timestamp. Whenever a track tree node accesses a new child node, the update timestamp of the node itself and all the direct ancestor nodes of the node are updated to the creation timestamp of the new child node. Among them, the update timestamp of the newly created node is equal to the creation timestamp. When traversing the matching forest, each trajectory tree is traversed first and foremost. When a timestamp difference of one of the nodes of the trajectory tree is greater than Q, and the node is not the root node of the trajectory tree, the tree branch corresponding to the node is pruned, where timestamp difference = current timestamp - update timestamp, the current timestamp refers to the frame sequence information corresponding to the binary image of the current frame when traversing the matching forest, and Q is the second preset threshold.

7. The method for detecting and tracking high-altitude objects according to claim 1 or 2, characterized in that: In the step S500, it is determined whether there is a trajectory path in the trajectory path corresponding to the trajectory tree to be deleted that meets the preset suspected parabolic condition. If there is a trajectory path that meets the preset suspected parabolic condition, the method of determining that the trajectory path is a suspected parabolic path specifically comprises the following operations: The trajectory path whose total number of nodes is higher than the preset node number threshold, whose acceleration of nodes along the way is lower than the preset acceleration threshold, whose angular velocity of nodes along the way is lower than the preset angular velocity threshold, and whose deflection angle of nodes along the way is lower than the preset deflection angle threshold is selected as the suspected parabolic path.

Images