Dynamic Trajectory Fusion Method and System Based on Sliding Window

Abstract

The present invention discloses a dynamic trajectory fusion method and system based on a sliding window. The method includes: associating spatial target trajectories based on a trajectory cataloging method that fits the recently discovered point set using a sliding window to obtain a trajectory association result; filtering the trajectory association result based on a noise trajectory filtering method that suppresses trajectory dormancy to obtain target trajectories; supplementing and fusing blank points and blank segments of the target trajectories to obtain final target trajectories. The system includes: a trajectory association module, a trajectory filtering module, and a supplementing and fusing module. By using the present invention, stable and correct target trajectories can be obtained. As a dynamic trajectory fusion method and system based on a sliding window, the present invention can be widely applied to the field of trajectory fusion technology.

Description

Technical Field

[0001] The present invention relates to the technical field of trajectory fusion, and in particular, to a dynamic trajectory fusion method and system based on a sliding window. Background Art

[0002] For space targets, due to the large unit field of view of the observation instrument, the relatively small size of the space target and the very long distance, the target imaging is extremely weak. Also, since the results obtained by the front-end detection algorithm contain a large amount of noise, and due to the complexity of the space target imaging environment, there is generally a phenomenon of star occlusion during the target imaging process, resulting in missed detections or long interruptions in the front-end detection algorithm, causing interruptions in the trajectory association of the back-end trajectory association. Moreover, the target has the characteristics of global non-uniform non-linear and local approximately uniform linear motion, so it is difficult to obtain a stable and correct target trajectory.

[0003] Chinese Patent Document No. CN114236579A discloses a method and device for filtering drift points after GPS multi-trajectory fusion. This method judges whether a trajectory point is a valid point or a drift point by setting an authoritative point and calculating the distance between the trajectory point and the valid authoritative point. The valid points are added to the trajectory list, and the drift points, that is, the invalid position information, are filtered after being verified, realizing trajectory correction and improving the accuracy of the trajectory, but it is only effective for targets with uniform linear motion. Summary of the Invention

[0004] In order to solve the above technical problems, the object of the present invention is to provide a dynamic trajectory fusion method and system based on a sliding window, which can obtain a stable and correct target trajectory.

[0005] The first technical solution adopted by the present invention is: a dynamic trajectory fusion method based on a sliding window, comprising the following steps:

[0006] Associating the space target trajectories based on a trajectory cataloging method that fits the recently discovered point set using a sliding window to obtain a trajectory association result;

[0007] Filtering the trajectory association result based on a noise trajectory filtering method for trajectory dormancy suppression to obtain a target trajectory;

[0008] Supplementary fusion of the blank points and blank segments of the target trajectory to obtain the final target trajectory.

[0009] Further, the step of associating the space target trajectories based on a trajectory cataloging method that fits the recently discovered point set using a sliding window to obtain a trajectory association result specifically includes:

[0010] Obtaining the current window trajectory and historical trajectory and setting a trajectory distance threshold and a target distance threshold;

[0011] Perform linear fitting on the recently discovered point set of the historical trajectory to obtain a set of lines;

[0012] Select the most recently discovered points for the current window trajectory, calculate the distances from the most recently discovered points to all the lines in the line set, and obtain a point-line distance set;

[0013] Select the trajectory pairs with distances less than the trajectory distance threshold from the point-line distance set and set them as candidate catalog objects;

[0014] Select the points of the nearest frame of the historical trajectory in the trajectory pairs from the candidate catalog objects;

[0015] Calculate the distance between the most recently discovered point and the point of the nearest frame of the historical trajectory in the trajectory pair;

[0016] When the distance between the most recently discovered point and the point of the nearest frame of the historical trajectory in the trajectory pair is less than the target distance threshold, catalog the trajectory pair to obtain a trajectory association result.

[0017] Furthermore, the step of filtering the trajectory association result by the noise trajectory filtering method based on trajectory dormancy suppression to obtain the target trajectory specifically includes:

[0018] Obtain the current window trajectory association result and set a trajectory length threshold and a growth speed threshold;

[0019] Construct a trajectory history record buffer set;

[0020] Calculate the trajectory length and growth speed of the current window trajectory association result;

[0021] Eliminate the noise trajectories in the current window trajectory association result with a trajectory length less than the trajectory length threshold and a growth speed less than the growth speed threshold;

[0022] Store the trajectory association result after eliminating the noise trajectories into the trajectory history record buffer set to obtain the target trajectory.

[0023] Furthermore, the step of supplementing and fusing the blank points and blank segments of the target trajectory to obtain the final target trajectory specifically includes:

[0024] Supplement and fuse the blank points in the target trajectory by the multi-feature fusion blank point supplement method based on adaptive point set selection;

[0025] Supplement and fuse the blank segments in the target trajectory by the abnormal trajectory self-organization fusion algorithm based on the fusing mechanism;

[0026] Integrate the target trajectory after supplementing and fusing the blank points and blank segments to obtain the final target trajectory.

[0027] Further, the step of supplementing and fusing blank points in the target trajectory by the multi-feature fusion blank point supplement method based on adaptive point set selection specifically includes:

[0028] Select a point set for sampling according to the position where the blank point appears in the target trajectory to obtain a sample point set;

[0029] Fit the trajectory direction and trajectory speed of the sample point set;

[0030] Determine the position of the most suitable blank point according to the trajectory direction and trajectory speed of the sample point set, and supplement the blank point.

[0031] Further, the step of supplementing and fusing blank segments in the target trajectory by the abnormal trajectory self-organization fusion algorithm based on the fusing mechanism specifically includes:

[0032] Obtain the original trajectory and the current window trajectory, and set the fusing threshold;

[0033] Fill in blank points for the blank segments of the current window trajectory, and calculate the number of filled blank points;

[0034] When the number of filled blank points is greater than the fusing threshold, trigger fusing;

[0035] Fuse and correct the original trajectory and the current window trajectory based on the fusing mechanism.

[0036] The second technical solution adopted by the present invention is: a dynamic trajectory fusion system based on a sliding window, including:

[0037] A trajectory association module that associates the spatial target trajectory based on the trajectory cataloging method fitted by the recently discovered point set using a sliding window to obtain a trajectory association result;

[0038] A trajectory filtering module that filters the trajectory association result based on the noise trajectory filtering method of trajectory sleep suppression to obtain a target trajectory;

[0039] A supplementing and fusing module for supplementing and fusing blank points and blank segments of the target trajectory to obtain a final target trajectory.

[0040] The beneficial effects of the method and system of the present invention are as follows: First, based on a sliding window, a trajectory cataloging method that fits the recently discovered point set is used. By utilizing the motion characteristics of space targets, where the imaging in the camera is globally non-uniform and non-linear while locally approximately uniform and linear, trajectories can be stably associated. Then, a noise trajectory filtering method based on trajectory dormancy suppression is used to filter the trajectory association results, effectively filtering out noise trajectories to obtain target trajectories. Subsequently, for the situation where there are blank points and blank segments in the target trajectory due to occlusion during the target imaging process, blank point and blank segment supplementation and fusion are carried out through multi-feature fusion blank point supplementation based on adaptive point set selection, or by using an abnormal trajectory self-organization fusion algorithm based on a fusing mechanism, thereby obtaining a stable and correct target trajectory. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] Figure 1 is a flowchart of the steps of a dynamic trajectory fusion method based on a sliding window according to the present invention;

[0042] Figure 2 is a structural block diagram of a dynamic trajectory fusion system based on a sliding window according to the present invention;

[0043] Figure 3 is a schematic diagram of the motion characteristics of an on-orbit target in an image in a specific embodiment of the present invention;

[0044] Figure 4 is a schematic diagram of the imaging trajectory of a space target in a specific embodiment of the present invention;

[0045] Figure 5 is a schematic diagram of the target detection result containing noise in a specific embodiment of the present invention;

[0046] Figure 6 is a schematic diagram of the effect of a noise trajectory filtering algorithm based on trajectory dormancy suppression in a specific embodiment of the present invention;

[0047] Figure 7 is a schematic diagram of missed detection points in Case 1 in a specific embodiment of the present invention;

[0048] Figure 8 is a schematic diagram of missed detection points in Case 2 in a specific embodiment of the present invention;

[0049] Figure 9 is a schematic diagram of missed detection points in Case 3 in a specific embodiment of the present invention;

[0050] Figure 10 is a schematic diagram of the effect of a blank point supplementation algorithm in a specific embodiment of the present invention;

[0051] Figure 11 is a schematic diagram showing a significant difference between the actual position and the predicted position of a trajectory after a long-term missed detection interruption in a specific embodiment of the present invention;

[0052] Figure 12 This is a schematic diagram showing the effect of the abnormal trajectory self - organizing fusion algorithm based on the fusing mechanism in a specific embodiment of the present invention. Specific Embodiment

[0053] The following further elaborates on the present invention in detail in conjunction with the attached drawings and specific embodiments. For the step numbers in the following embodiments, they are only set for the convenience of elaboration and explanation, and no limitation is imposed on the order between steps. The execution order of each step in the embodiments can be adaptively adjusted according to the understanding of those skilled in the art.

[0054] Referring to Figure 1 , the present invention provides a dynamic trajectory fusion method based on a sliding window, and this method includes the following steps:

[0055] S1. Use a trajectory cataloging method that fits the recently discovered point set based on a sliding window to associate the space target trajectories, and obtain the trajectory association result;

[0056] Specifically, as Figure 3 shown, the imaging of the on - orbit target in the camera is globally non - uniform and non - linear, and locally approximately uniform and linear. Therefore, during the process of the sliding window, when cataloging the trajectory generated by each new window with the historical trajectory, the linear fitting of the historical trajectory needs to use the "recently discovered point set" and should not use the entire point set of the trajectory.

[0057] Among them, the recently discovered point set is to expand and select the current point, and select the 5 points closest to the current point on the same trajectory.

[0058] Also, due to situations such as being too close, relative motion, crossing, intersection, etc. existing between the trajectories of the space target imaging, as Figure 4 shown, in order to distinguish different trajectories, different shapes are used to represent the trajectory points. The trajectory formed by the triangular trajectory points indicated by the arrow overlaps and intersects with the trajectory formed by the circular trajectory points. To handle such situations, the trajectory cataloging method based on the fitting of the recently discovered point set needs to be divided into two steps.

[0059] S1.1. Select candidate cataloging objects;

[0060] Specifically, first obtain the current window trajectory T = {t 1 , t 2 ,..., t m} and the historical trajectory H = {h 1 , h 2 ,... h n} and set the trajectory distance threshold Thd1 and the target distance threshold Thd2; then traverse the historical trajectory, and use the least - squares method to perform linear fitting on the recently discovered point set of each historical trajectory to obtain the line set L = {l i}; Then traverse the trajectories found in the current window. For each trajectory, select the latest discovered point M 0 , calculate the distance from M 0 to all the lines in L. Assume the direction vector of line l i is s, then the distance d 0 from M i to l l is d l ; Finally, set the trajectory pairs with d l less than the trajectory distance threshold Thd1 as the candidate catalog objects C = {c i}.

[0061] Among them, the specific calculation formula of d l is as follows:

[0062]

[0063] In the above formula, M is, is.

[0064] S1.2. Select the most suitable trajectory pair from the candidate catalog objects for cataloging to obtain the trajectory association result.

[0065] Specifically, first traverse the candidate catalog objects and select the point M 1 of the nearest frame of the historical trajectory in the trajectory pair; then calculate the distance d 0 between M 1 and M p , and determine whether it conforms to the speed distribution of the current historical trajectory according to d p . If d p is less than the target distance threshold Thd2, it conforms to the speed distribution of the current historical trajectory, and catalog the conforming trajectory pairs; if d p is equal to or greater than the target distance threshold Thd2, re-select the point of the nearest frame of the historical trajectory in the trajectory pair for distance calculation.

[0066] Through the above method, not only can the targets of uniform linear motion be excluded, but also the targets of global non-uniform non-linear and local approximately uniform linear motion can be stably associated with trajectories, and it has strong robustness against the interference of situations such as trajectories being too close, relative motion, crossing, intersection, etc.

[0067] S2. Filter the trajectory association result by the noise trajectory filtering method based on trajectory dormancy suppression to obtain the target trajectory;

[0068] Specifically, due to the extremely low signal-to-noise ratio of the target, in order to ensure the detection of the target as much as possible, the relevant thresholds for target detection are set very low. This results in introducing a large amount of noise into the detection results while detecting the target as much as possible. When these detection results containing a large amount of noise are input into the trajectory association module, it brings great interference to the trajectory association. When the trajectory association algorithm associates the correct trajectory, it will also associate many noise trajectories that meet the trajectory constraint characteristics, such as Figure 5 As shown, the arrow indicates the target trajectory, and the rest are noise trajectories.

[0069] From Figure 5 it can be seen that these noise trajectories seem chaotic, but the associated trajectories are also in uniform straight lines. However, the noise trajectories have two characteristics: one is that the noise trajectories grow slowly or even stagnate; the other is that the absolute length of the noise trajectories is relatively short. According to these two characteristics, the sleep suppression mechanism can be used to judge and eliminate the noise trajectories.

[0070] First, when a trajectory is first associated, the trajectory type is not confirmed. Instead, after a certain length of window sliding, the trajectory length and growth rate of the current window trajectory association result are calculated. Secondly, set the trajectory length threshold and growth rate threshold and construct a trajectory history record buffer set. Then, make a comparison. When the trajectory length in the current window trajectory association result is less than the trajectory length threshold and the growth rate is less than the growth rate threshold, it is determined that the trajectory has entered the sleep state, suppress the trajectory in the sleep state, determine it as a noise trajectory, and eliminate the noise trajectory in the current window trajectory association result. Finally, store the trajectory association result after eliminating the noise trajectory into the trajectory history record buffer set to obtain the target trajectory, realizing the confirmation of the target trajectory and the filtering of the noise trajectory.

[0071] After filtering the noise trajectories using the above method, the effect is as shown in Figure 6 As shown in the figure, it can be seen that this mechanism has strong robustness to the detection results with extremely low signal-to-noise ratio and can effectively filter noise trajectories.

[0072] S3. Supplement and fuse the blank points and blank segments of the target trajectory to obtain the final target trajectory.

[0073] S3.1. Supplement and fuse the blank points in the target trajectory using the multi-feature fusion blank point supplement method based on adaptive point set selection;

[0074] Specifically, due to the complexity of the spatial target imaging environment, the phenomenon of star occlusion generally exists during the target imaging process, resulting in missed detection by the front-end detection algorithm and causing the interruption of trajectory association in the back-end trajectory association. At this time, it is necessary to supplement the missed points with blank points.

[0075] According to the imaging characteristics of the target motion, the target has the characteristics of global non-uniform and non-linear motion and local approximately uniform linear motion. Therefore, when sampling and fitting the current trajectory point, it is necessary to select the adjacent point set. According to the position of the missed detection point in the trajectory, there are three cases in total, and the sampling methods for each case are different. The specific analysis is as follows:

[0076] Case 1: The missed detection point appears at the front end of the trajectory. As Figure 7 shown, at this time, the nearest neighbor sampling point needs to be selected from the point set after the missed detection point.

[0077] Case 2: The missed detection point appears in the middle of the trajectory. As Figure 8 shown, at this time, the nearest neighbor sampling point needs to be selected from the point sets before and after the missed detection point simultaneously.

[0078] Case 3: The missed detection point appears at the rear end of the trajectory. As Figure 9 shown, at this time, the nearest neighbor sampling point needs to be selected from the point set before the missed detection point.

[0079] After determining the situation where the missed detection point appears, adaptively select the point set for sampling, and then use the least squares fitting line method to fit the trajectory direction and trajectory speed of the sample point set respectively. Finally, determine the most suitable position of the missed detection point to supplement the missed detection point.

[0080] Perform blank point supplementation according to the above blank point supplementation algorithm, and its effect is as Figure 10 shown. The circles in the left figure are the situations where missed detection interruptions occur, and the right figure makes correct supplements at the corresponding positions. It can be seen from the figure that for missed detections caused by situations such as stellar occlusion, the missed detection points can still be supplemented, and the trajectory can be stably associated.

[0081] It should be noted that since the effects of different algorithms will interact and affect each other, when each new algorithm module is added, there will be a slight difference in the final processing effect of the noise trajectory. When all algorithm modules are in place, the best effect will be achieved. S3.2. The self-organizing fusion algorithm for abnormal trajectories based on the fusing mechanism supplements and fuses the blank segments in the target trajectory;

[0082] Specifically, as Figure 11 shown, due to the widespread occlusion in the imaging of space targets, when the occlusion situations occur densely, it will lead to long-term interruptions of the trajectory. Although the blank point supplementation algorithm in step S3.1 can effectively handle short-term missed detection interruptions, the target trajectory is not a global uniform linear motion. Therefore, when long-term missed detection interruptions occur, the supplemented blank points will have a significant difference from the actual target trajectory, resulting in the situation of "trajectory drift". The trajectory segment detected by the new window again will be recognized as a different trajectory from the historical segment.

[0083] To solve the above situation, a blank point supplement fuse mechanism is introduced, and an abnormal trajectory self-organizing fusion method based on the fuse mechanism is adopted. When a trajectory with long-term undetected occurs, the trajectory state is marked as "deficient". When a fragment of the same trajectory is detected again, due to the blank point supplement mechanism, a large number of blank points will be supplemented to the new trajectory fragment in a short time. The trajectory generated by this fragment is regarded as an abnormal trajectory, thereby triggering the fuse mechanism to self-organize and fuse the abnormal trajectory and the marked long-term interrupted trajectory.

[0084] As Figure 12 shown, to distinguish different trajectories, different shapes are used to represent trajectory points. In Fig. (a), the trajectory formed by triangular trajectory points represents the original trajectory, and the trajectory formed by circular trajectory points represents the current window trajectory. The trajectory in Fig. (c) is the final target trajectory after fusion. Since the original trajectory with blank points supplemented due to long-term undetected and the newly discovered abnormal trajectory, i.e., the current window trajectory, has filled a large number of blank points at this time, most of the points of the two trajectories are linearly distributed. All points can be used for linear fitting to find the intersection point. According to the distribution of points near the intersection point of the two trajectories, the point closest to the intersection point is selected as the best fusion point for the original trajectory and the current window trajectory respectively. The trajectory before and including the intersection point of the original trajectory is intercepted, and the trajectory before and including the intersection point of the current window trajectory is intercepted. The two intercepted trajectories are fused to form a new correct trajectory, completing the fusion correction.

[0085] Through the above method, for different fragments of the same trajectory that reappear after being interrupted due to long-term occlusion and show obvious offsets, they can be self-organized for re-cataloging and fusion.

[0086] S3.2.1. Obtain the original trajectory and the current window trajectory, and set the fuse threshold;

[0087] Specifically, assume that the original trajectory is the trajectory marked in the "deficient" state, and the trajectory t j , and the current window trajectory C = {c 1 , c 2 ,..., c n}, and the fuse threshold is Thd.

[0088] S3.2.2. Fill the blank segments of the current window trajectory with blank points, and calculate the number of filled blank points;

[0089] S3.2.3. When the number of filled blank points is greater than the fuse threshold, trigger the fuse, and perform fusion correction on the original trajectory and the current window trajectory based on the fuse mechanism.

[0090] Specifically, when the number of filled blank points is greater than the fuse threshold Thd, tj Fuse and correct with c i Remove the "deficiency" mark of t j and delete c i .

[0091] S3.3. Integrate and supplement the target trajectory after filling in the blank points and blank segments of the fused trajectory to obtain the final target trajectory.

[0092] As Figure 2 shown, a dynamic trajectory fusion system based on a sliding window includes:

[0093] A trajectory association module that associates spatial target trajectories based on a trajectory cataloging method that fits the recently discovered point set using a sliding window to obtain a trajectory association result;

[0094] A trajectory filtering module that filters the trajectory association result based on a noise trajectory filtering method that suppresses trajectory dormancy to obtain a target trajectory;

[0095] A supplementary fusion module for supplementing and fusing the blank points and blank segments of the target trajectory to obtain the final target trajectory.

[0096] The content in the above method embodiments is applicable to the system embodiments. The functions specifically implemented by the system embodiments are the same as those of the above method embodiments, and the beneficial effects achieved are also the same as those of the above method embodiments.

[0097] The above is a specific description of the preferred embodiments of the present invention. However, the present invention is not limited to the described embodiments. Those skilled in the art can make various equivalent deformations or substitutions without departing from the spirit of the present invention, and these equivalent deformations or substitutions are all included within the scope defined by the claims of this application.

Claims

1. A dynamic trajectory fusion method based on a sliding window, characterized in that, it includes the following steps: Associate the space target trajectory based on the trajectory cataloging method that fits the recently discovered point set using a sliding window to obtain a trajectory association result; Filter the trajectory association result based on the noise trajectory filtering method of trajectory dormancy suppression to obtain the target trajectory; Supplement and fuse the blank points and blank segments of the target trajectory to obtain the final target trajectory; Among them, the step of associating the space target trajectory based on the trajectory cataloging method that fits the recently discovered point set using a sliding window to obtain a trajectory association result specifically includes: Obtain the current window trajectory and the historical trajectory and set the trajectory distance threshold and the target distance threshold; Perform linear fitting on the recently discovered point set of the historical trajectory to obtain a set of lines; Select the most recently discovered point for the current window trajectory, calculate the distances from the most recently discovered point to all the lines in the set of lines to obtain a set of point-line distances; Select the trajectory pairs with distances less than the trajectory distance threshold from the set of point-line distances as candidate cataloging objects; Select the points of the most recent frame of the historical trajectory in the trajectory pairs from the candidate cataloging objects; Calculate the distance between the most recently discovered point and the points of the most recent frame of the historical trajectory in the trajectory pair; When the distance between the most recently discovered point and the points of the most recent frame of the historical trajectory in the trajectory pair is less than the target distance threshold, catalog the trajectory pair to obtain a trajectory association result; The step of supplementing and fusing the blank points and blank segments of the target trajectory to obtain the final target trajectory specifically includes: Supplement and fuse the blank points in the target trajectory based on the multi-feature fusion blank point supplement method of adaptive point set selection; Supplement and fuse the blank segments in the target trajectory based on the abnormal trajectory self-organization fusion algorithm based on the fusing mechanism; Integrate the target trajectory after supplementing and fusing the blank points and blank segments to obtain the final target trajectory; The step of supplementing and fusing the blank segments in the target trajectory based on the abnormal trajectory self-organization fusion algorithm based on the fusing mechanism specifically includes: Obtain the original trajectory and the current window trajectory and set the fusing threshold; Fill in blank points for the blank segments of the current window trajectory and calculate the number of filled blank points; When the number of filled blank points is greater than the fusing threshold, trigger fusing; Fuse and correct the original trajectory and the current window trajectory based on the fusing mechanism.

2. The dynamic trajectory fusion method based on a sliding window according to claim 1, characterized in that, the step of filtering the trajectory association result based on the noise trajectory filtering method of trajectory dormancy suppression to obtain the target trajectory specifically includes: Obtain the current window trajectory association result and set the trajectory length threshold and the growth speed threshold; Construct a trajectory history record buffer set; Calculate the trajectory length and growth speed of the current window trajectory association result; Remove the noise trajectories in the current window trajectory association result with a trajectory length less than the trajectory length threshold and a growth speed less than the growth speed threshold; Store the trajectory association result after removing the noise trajectories into the trajectory history record buffer set to obtain the target trajectory.

3. The dynamic trajectory fusion method based on a sliding window according to claim 1, characterized in that, The step of supplementing and fusing the blank points in the target trajectory by the multi-feature fusion blank point supplement method based on adaptive point set selection specifically includes: Select a point set for sampling according to the positions where the blank points appear in the target trajectory to obtain a sample point set; Fit the trajectory direction and trajectory speed of the sample point set; Determine the positions of the most suitable blank points according to the trajectory direction and trajectory speed of the sample point set, and supplement the blank points.

4. A dynamic trajectory fusion system based on a sliding window, characterized in that, it includes: A trajectory association module that associates the space target trajectories based on a trajectory cataloging method that fits the recently discovered point set using a sliding window to obtain a trajectory association result; A trajectory filtering module that filters the trajectory association result based on a noise trajectory filtering method for trajectory dormancy suppression to obtain a target trajectory; A supplement and fusion module for supplementing and fusing the blank points and blank segments of the target trajectory to obtain a final target trajectory; Among them, the step of associating the space target trajectories based on a trajectory cataloging method that fits the recently discovered point set using a sliding window to obtain a trajectory association result specifically includes: Obtain the current window trajectory and historical trajectories and set a trajectory distance threshold and a target distance threshold; Perform linear fitting on the recently discovered point set of the historical trajectories to obtain a set of lines; Select the most recently discovered point for the current window trajectory, calculate the distances from the most recently discovered point to all the lines in the set of lines to obtain a point-line distance set; Select the trajectory pairs with distances less than the trajectory distance threshold from the point-line distance set as candidate cataloging objects; Select the points of the most recent frame of the historical trajectory in the trajectory pairs from the candidate cataloging objects; Calculate the distance between the most recently discovered point and the points of the most recent frame of the historical trajectory in the trajectory pair; When the distance between the most recently discovered point and the points of the most recent frame of the historical trajectory in the trajectory pair is less than the target distance threshold, catalog the trajectory pair to obtain a trajectory association result; The step of supplementing and fusing the blank points and blank segments of the target trajectory to obtain a final target trajectory specifically includes: Supplement and fuse the blank points in the target trajectory by the multi-feature fusion blank point supplement method based on adaptive point set selection; Supplement and fuse the blank segments in the target trajectory by the abnormal trajectory self-organization fusion algorithm based on a fusing mechanism; Integrate the target trajectory after supplementing and fusing the blank points and blank segments to obtain a final target trajectory; The step of supplementing and fusing the blank segments in the target trajectory by the abnormal trajectory self-organization fusion algorithm based on a fusing mechanism specifically includes: Obtain the original trajectory and the current window trajectory, and set a fusing threshold; Fill in blank points for the blank segments of the current window trajectory and calculate the number of filled blank points; When the number of filled blank points is greater than the fusing threshold, trigger fusing; Fuse and correct the original trajectory and the current window trajectory based on the fusing mechanism.

Images