A target detection method based on ground-based telescope data

Abstract

The present application belongs to the technical field of space target detection, and discloses a target detection method based on ground-based telescope data, comprising the following steps: reading all the time-series images of ground-based telescopes under the input folder path and performing pretreatment; identifying and extracting space satellite targets in the pretreated images; performing forward inter-frame target correlation identification on the identified and extracted space satellite targets according to time sequence, extracting the motion trajectory of the moving target, and eliminating the false target points; after the forward inter-frame target correlation identification on the identified and extracted space satellite targets according to time sequence, performing reverse inter-frame target correlation identification according to time sequence, and then merging the forward correlation identification result and the reverse correlation identification result of each frame. The method can reduce the false alarm rate, improve the detection capability and work efficiency in the process of space target detection.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of space target detection, and particularly relates to a target detection method based on ground-based telescope data. BACKGROUND

[0002] The space target detection methods mainly include photoelectric detection systems and radar detection systems. The photoelectric detection system has the characteristics of low cost, low energy consumption, high sensitivity and long detection distance, and is widely favored by countries because it uses a short observation band and detects target reflection or radiation information. At present, the detection and cataloging of low-orbit space targets has been relatively complete, and radar and photoelectric equipment are used for simultaneous detection. The photoelectric detection system can assist in improving the accuracy of the radar system. The detection of small-size space targets and medium-high-orbit space targets mainly relies on photoelectric detection systems. In order to improve the search efficiency of the photoelectric detection system, the equipment is developing towards large aperture and large field of view. The purpose of large aperture is to improve the light throughput, greatly shorten the exposure time, improve the image signal-to-noise ratio, and improve the detection ability of dark space targets. The purpose of large field of view is to improve the detection range, so that the telescope can cover more sky areas in the same time. Ground-based telescopes are an important part of the space target wide-area monitoring system, and have carried out many researches.

[0003] Due to the limited observation resources, it is difficult to meet the monitoring of a large number of space debris. There are three important development trends for space target detection technology based on ground-based telescopes: extending the working time of ground-based telescopes, improving the image quality of ground-based telescopes, and improving the detection ability and working efficiency of space target detection algorithms.

[0004] Improving the detection ability and working efficiency of space target detection algorithms: on the one hand, small-size space debris in medium-high orbits are far away and have low brightness, and the noise in the image is difficult to distinguish from dark space targets, which easily causes a large number of false alarm points; on the other hand, there are a large number of background stars in the field of view, and the existence of background stars causes a large number of false alarm points, which greatly interferes with space target detection. These factors comprehensively lead to an increase in the false alarm rate, a decrease in the detection ability and working efficiency in the space target detection process. SUMMARY

[0005] The purpose of the application is to provide a target detection method based on ground-based telescope data, which can easily distinguish noise from dark space targets in the image during space target detection, remove the interference of background stars, reduce the false alarm rate, and improve the detection ability and working efficiency.

[0006] The technical scheme adopted by the application is a target detection method based on ground-based telescope data, which includes the following steps:

[0007] S1, read all ground-based telescope time-series images under the input folder path and pre-process them;

[0008] S2, identify and extract space satellite targets from the pre-processed images;

[0009] S3, perform forward inter-frame target association identification on the identified and extracted space satellite targets according to time series, extract the motion trajectory of the moving target, and eliminate false target points;

[0010] S4, after performing forward inter-frame target association identification on the identified and extracted space satellite targets according to time series, perform reverse inter-frame target association identification according to time series, and combine the forward and reverse association identification results of each frame.

[0011] Further, in S1, the pre-processing specifically includes:

[0012] correcting the uneven background of single-frame image data, and performing fast registration on all image data based on the gray-scale projection method.

[0013] Further, in S2, the identification and extraction of space satellite targets from the pre-processed images specifically includes:

[0014] (1) extract all space targets in the image by calculating an adaptive threshold, and perform denoising and edge optimization on the extraction results;

[0015] (2) remove the interference of star targets in all extracted space targets using the time-series image inter-frame difference method, and identify and extract space satellite targets by combining morphological image processing algorithms.

[0016] Further, in S3, the forward inter-frame target association identification of the identified and extracted space satellite targets according to time series specifically includes:

[0017] (1) determine the search starting point;

[0018] (2) according to the determined starting trajectory, select two target points of one of the trajectories to predict the target position in the third frame;

[0019] (3) start the next process, search from the second frame, all coordinate points in the second frame are starting points, refer to steps (1) to (2), until all coordinate points in the Nth frame are starting points;

[0020] (4) delete the repeated false alarm tracks.

[0021] Further, the determination of the search starting point specifically includes:

[0022] Start searching from the first frame, all coordinate points of targets in the first frame are starting points, set a speed window L1 according to the exposure time length, P1 is a target in the first frame, search for a target in the second frame falling into the window in the speed window of P1:

[0023] If no target falls into the second frame, prune and delete the track, and start searching from other target points in the first frame;

[0024] If P2 exists as a target in the second frame and falls into the speed window L1 of P1, all target points in the first frame and all required combinations of starting points falling into the speed window L1 are recorded as starting tracks.

[0025] Further, according to the determined starting track, two target points of one of the tracks are selected to predict the target position in the third frame, specifically:

[0026] According to the starting track determined by the target points in the first frame and the second frame and the time sequence time, two target points of one of the tracks are selected to predict the position of the target P3 in the third frame, and the detection window L2 radius in the nth frame Wherein, n≥3, ε is a system error, and there are three cases during detection:

[0027] If no target point falls into the detection window L2, prune and delete the track, terminate the depth-first search starting from the target P1 in the first frame, and start the depth-first search from other tracks, and the process is until the Nth frame, N is the total number of frames;

[0028] If only one target point falls into the detection window L2, the point is counted into the track, then the least square method is used to predict the target point position in the next frame using the updated track data and calculate the detection window size of the next frame;

[0029] If multiple target points fall into the detection window L2, the center point of the spatial target is estimated using the first moment, the point is counted into the track, and then the least square method is used to predict the target point position in the next frame using the updated track data and calculate the detection window size of the next frame.

[0030] Further, the repeated false alarm tracks are deleted, and specifically:

[0031] In multiple tracks of the same starting point, the track curve is fitted by using the track position, and the error of each track is calculated by using the associated target point position and the track curve, the track with the minimum error is retained, and the other tracks are deleted.

[0032] The beneficial effects of the present application are:

[0033] (1) Reduce the false alarm rate.

[0034] (2) Improve detection ability and working efficiency. Attached Figure Description

[0035] Figure 1 This is a flowchart of the present invention;

[0036] Figure 2 This is a flowchart of the inter-frame target association recognition process of the present invention;

[0037] Figure 3 This is a schematic diagram illustrating the process of determining the search starting point in this invention;

[0038] Figure 4 A schematic diagram of the flight paths triggered by false alarm points;

[0039] Figure 5 This is a schematic diagram illustrating the forward association recognition and reverse association recognition of the present invention;

[0040] Figure 6 The results of the test using the method of this invention are shown in the figure.

[0041] Wherein, L1 - velocity window; L2 - detection window; p1 - target in the first frame; p2 - target in the second frame; p3 - target in the third frame; S - trajectory; A - real target; B - false alarm point; C - target; D - missing. Detailed Implementation

[0042] To make the objectives, technical solutions, and advantages of this invention clearer and more understandable, the technical solutions of this invention will be clearly and completely described below in conjunction with the accompanying drawings.

[0043] like Figure 1 As shown, the target detection method based on ground-based telescope data of the present invention includes the following steps:

[0044] S1, Read all ground-based telescope time-series images under the input folder path and perform preprocessing;

[0045] First, based on the input folder path, all time-series images of ground-based telescopes under the path are read to obtain the grayscale matrix data and header information of each frame image. Then, the non-uniform background of the single frame image data is corrected, and all image data are quickly registered based on the grayscale projection method to complete the image preprocessing.

[0046] S2, identifying and extracting space satellite targets from the preprocessed image, includes the following steps:

[0047] (1) Extract all spatial targets in the image by calculating an adaptive threshold, and perform denoising and edge optimization on the extraction results;

[0048] (2) In all extracted space targets, the time sequence image frame difference method is used to remove the interference of star targets, and a morphological image processing algorithm is combined to identify and extract space satellite targets.

[0049] S3, the space satellite targets identified and extracted are positively correlated and identified according to time sequence, the motion trajectory of the moving target is extracted, and the false target points are removed;

[0050] The present application intends to use the method of straight line detection to correlate and identify the space satellite targets identified and extracted, and the motion of the space satellite targets in a short time can be considered as uniform linear motion, and the motion parameters need to be estimated, and the least square method is an optimal parameter estimation method for uniform linear motion, but in the case of large amount of data, the dimension disaster will occur due to too many invalid false alarm points, and an effective solution is to prune the data in the depth first search process, and the data not in the speed window is pruned and does not participate in the least square estimation operation, and the wrong track is also pruned, so that the dimension disaster is avoided, and the operation capacity is greatly improved under the condition of high false alarm, such as Figure 2 、 Figure 3 As shown in the figure, the specific steps are as follows:

[0051] (1) determining the search starting point;

[0052] The process of determining the search starting point is also the process of determining the starting track, specifically, searching from the first frame, the coordinate points of all targets (noise, lens stain, false alarm target, real target, etc.) in the first frame are starting points, the speed window L1 is set according to the exposure time, P1 is a target in the first frame, and the target in the second frame falling into the window is searched in the speed window of P1:

[0053] If there is no target falling into the second frame, the track is pruned and deleted, and the search starts from other target points in the first frame.

[0054] If there is P2 as a target in the second frame (also a target in the first frame) and falling into the speed window L1 of P1, all the target points in the first frame and the starting points falling into the speed window L1 are recorded as starting tracks.

[0055] (2) according to the determined starting track, selecting two target points of one track to predict the target position in the third frame;

[0056] Specifically, according to the starting track determined by the target points in the first frame and the second frame and the time sequence time, one track is selected, such as the two target points of the track S to predict the position of the target P3 (P3 is also a target in the second frame and the third frame) in the third frame, and the detection window L2 radius Where, ε is the system error, there are three cases when detecting:

[0057] If there is no target point falling into the detection window L2, the track is pruned and deleted, the depth-first search starting from the first frame target P1 is terminated, and the depth-first search is started from other tracks, and this process continues until the Nth frame, N is the total number of frames.

[0058] If only one target point falls into the detection window L2, the point is counted into the track, and then the least square method is used to predict the next frame target point position and calculate the next frame detection window size using the updated track data.

[0059] If there are multiple target points falling into the detection window L2, this case does not exist most of the time, because usually the target is much larger than the detection window L2, it is difficult to appear this case, if it appears, the first moment estimation is used to estimate the center point of the space target, the point is counted into the track, and then the least square method is used to predict the next frame target point position and calculate the next frame detection window size using the updated track data.

[0060] (3) Start the next process, search from the second frame, all coordinate points in the second frame are starting points, refer to steps (1) to (2), until all coordinate points in the Nth frame are starting points.

[0061] (4) Delete the repeated false alarm tracks;

[0062] After the track is obtained by correlation recognition, a part of the false alarm points are too close to the track and induce the generation of false alarm tracks, as shown in Figure 4 , leading to the appearance of 2 or even multiple false alarm tracks for a starting point, most of the tracks of these false alarm tracks are the same, only the individual starting points are different, therefore after obtaining the track, the tracks with the same starting point are fused and deleted. Specifically, in the multiple tracks with the same starting point, the track curve is fitted using the track position, and the error of each track is calculated using the correlated target point position and the track curve, the track with the smallest error is retained, and the other tracks are deleted.

[0063] S4, after the space satellite target extracted by recognition is positively correlated and recognized between frames in time sequence, the space satellite target is negatively correlated and recognized between frames in time sequence, and then the positively correlated and recognized result and the negatively correlated and recognized result of each frame are combined;

[0064] As shown in Figure 5As shown, in the process of forward inter-frame target correlation identification in time sequence, the measured target can be missing in a frame image due to various reasons, and after the forward inter-frame target correlation identification in time sequence, the reverse inter-frame target correlation identification in time sequence is performed, and the forward correlation result and the reverse correlation result of each frame are combined, so that the defect of target detection failure caused by target loss in the forward correlation identification can be made up.

[0065] In a specific embodiment, MATLAB software coding is used for spatial satellite target detection of multi-frame ground-based telescope time sequence images, and the finally detected spatial satellite target is marked with a marked frame, and the effect is as shown in Figure 6

[0066] The contents not described in detail in the specification of the present application belong to the known technology of those skilled in the art.​

Claims

1. A target detection method based on ground-based telescope data, characterized in that, Includes the following steps: S1, Read all ground-based telescope time-series images under the input folder path and perform preprocessing; S2, identify and extract space satellite targets from the preprocessed image; S3, perform inter-frame target association identification on the identified and extracted space satellite targets in a forward time series, extract the motion trajectory of moving targets, and eliminate false target points; S4. After performing inter-frame target association identification on the extracted space satellite targets in the forward direction according to the time series, the inter-frame target association identification is then performed in the reverse direction according to the time series. Finally, the forward association identification results and the reverse association identification results of each frame are merged.

2. The target detection method based on ground-based telescope data according to claim 1, characterized in that, In S1, the preprocessing specifically includes: The non-uniform background of single-frame image data is corrected, and all image data are quickly registered based on the grayscale projection method.

3. The target detection method based on ground-based telescope data according to claim 1, characterized in that, In step S2, the identification and extraction of space satellite targets from the preprocessed image includes the following steps: (1) Extract all spatial targets in the image by calculating an adaptive threshold, and perform denoising and edge optimization on the extraction results; (2) Among all the extracted space targets, the interference of star targets is removed by the time-series image frame difference method, and the space satellite targets are identified and extracted by combining the morphological image processing algorithm.

4. The target detection method based on ground-based telescope data according to claim 1, characterized in that, In step S3, the identified and extracted space satellite targets are subjected to inter-frame target association identification in a forward time series. Specific steps include: (1) Determine the starting point of the search; (2) Based on the determined starting trajectory, select two target points on one of the trajectories to predict the target position in the third frame; (3) Start the next process, starting the search from the second frame. All coordinate points in the second frame are the starting point. Refer to steps (1) to (2) until all coordinate points in the Nth frame are the starting point. (4) Delete duplicate false alarm tracks.

5. The target detection method based on ground-based telescope data according to claim 4, characterized in that, The specific steps to determine the starting point of the search are: The search begins from the first frame, with the coordinates of all targets in the first frame serving as the starting point. A speed window L1 is set based on the exposure time, and P1 represents the targets in the first frame. The search then continues within the speed window of P1 to find targets that fall within the window in the second frame. If no target falls in the second frame, prune the trail, delete the trajectory, and start searching for other target points from the first frame. If P2 is a target in the second frame and falls within the velocity window L1 of P1, record all the required combinations of all target points in the first frame and the starting points falling within the velocity window L1 as the starting trajectory.

6. The target detection method based on ground-based telescope data according to claim 5, characterized in that, Based on the determined initial trajectory, the target position in the third frame is predicted by selecting two target points from one of the trajectories as follows: Based on the starting trajectory and timing determined from the target points in the first and second frames, two target points from one of the trajectories are selected to predict the position of target P3 in the third frame. The radius of the detection window L2 in the nth frame is also considered. Where n≥3, ε is the systematic error, and there are three cases during detection: If no target point falls within the detection window L2, the trajectory is pruned and deleted, the depth-first search of target P1 starting in the first frame is terminated, and the depth-first search starts from other trajectories. This process continues until the Nth frame, where N is the total number of frames. If only one target point falls within the detection window L2, then this point is included in the trajectory. Then, the updated trajectory data is used to predict the target point position in the next frame using the least squares method and to calculate the detection window size in the next frame. If multiple target points fall within the detection window L2, the center point of the spatial target is estimated using the first moment and included in the trajectory. Then, the target point position in the next frame is predicted using the least squares method and the detection window size in the next frame is calculated using the updated trajectory data.

7. The target detection method based on ground-based telescope data according to claim 4, characterized in that, The specific steps for deleting duplicate false alarm tracks are as follows: Among multiple tracks starting from the same point, the trajectory curve is fitted using the track position, and the error of each track is calculated using the associated target point position and trajectory curve. The track with the smallest error is retained, and the other tracks are deleted.

Images