Space target rapid association and extraction system and method

Abstract

The invention discloses a space target rapid association and extraction system. The system comprises a target point receiving module, a feature star map construction module, an initial track association module and a suspected / real track expansion module. The target point receiving module receives the star point target information and caches the star point target information; the feature star map construction module receives star point target information and constructs a feature star map; the initial trajectory association module performs inter-frame suspected trajectory association according to the star point target information and the feature star map to obtain an initial suspected trajectory, and outputs the initial suspected trajectory to a suspected / real trajectory expansion module; and the suspected / real trajectory expansion module expands the initial suspected trajectory into a real trajectory according to the feature star map. The invention further discloses a space target rapid association and extraction method. The method can effectively detect and track the space target and the motion trail characteristics of the space target, is high in calculation parallelism degree and low in calculation delay, and has good practicability and popularization value.

Description

Technical Field

[0001] The invention relates to a space target rapid association and extraction system and method, and belongs to the field of high performance computing technology in the field of computer technology. Background Art

[0002] With the continuous expansion of the scale of human development and utilization of space technology, space has become an important platform for obtaining and transmitting information, and the utilization and control of space information is particularly important. The premise of obtaining space information is to effectively detect, identify and track space targets. Space targets mainly refer to space debris, artificial satellites and various flying objects entering outer space. Space target detection, tracking and identification refers to detecting the required target from the complex starry sky background and obtaining the target's motion trajectory. It is an important part of space situational awareness and a key technical guarantee for ensuring the safety of manned space flight and space early warning. It has very important practical significance and application value.

[0003] Traditional space target detection faces many problems and challenges: first, the interference of sky light conditions and radiation noise in space images causes the background fluctuation of the captured images; second, the uneven camera process will lead to uneven images; third, the existence of dense star background in the field of view affects the efficiency and accuracy of target extraction; fifth, the signal-to-noise ratio of space targets is low, and target information and features are easily submerged by noise. The above characteristics have brought huge challenges to the extraction of space targets.

[0004] Space targets are small in size and weak in energy. They are imaged as point targets that occupy only a few pixels on the focal plane of ground-based / space-based telescopes and are easily submerged by background and noise. Traditional single-frame image processing has made it difficult to effectively detect and segment targets, and there are problems such as low detection rate and high false alarm rate. Therefore, it is an effective means to complete the trajectory association and target extraction of space targets based on the motion information of the target in the sequence frames.

[0005] The multi-frame detection methods of space targets are mainly divided into two categories: detection before tracking (DBT) and tracking before detection (TBD). DBT first processes a single frame image to determine whether there is a target in the single frame, and then uses the motion characteristics of the target in the sequence frame to confirm and correct the single frame detection results, and finally outputs the detection results to obtain the target. TBD first tracks possible targets in the image, and then establishes criteria based on the target characteristics, screens the target track, gradually eliminates false tracks, and finally obtains the correct target track.

[0006] The tracking-before-detection algorithm uses the continuity and similarity of the target for detection. It is difficult to handle targets that move quickly, and has high requirements for computing and storage resources, so it is not suitable for large-scale engineering projects. In contrast, the detection-before-tracking technology is a small target extraction algorithm with a wide range of applications. It has the advantages of simple calculation and fast operation, and is widely used in engineering practice.

[0007] The detection-before-tracking technology is divided into pre-processing and post-processing. The pre-processing stage obtains the star point data in the image through filtering, interpolation and background removal. The star point data includes both star points and space target points. The post-processing stage extracts the space target points in each frame of the image through motion analysis of the star point data between multiple frames, constructs the motion trajectory of the space target points, and realizes the detection and tracking of space targets.

[0008] When the image size is small, there are fewer star points and target points in each frame, and the traditional serial processor is used to complete the association and extraction of space targets. However, the small image size also limits the ability to perceive space targets by relying on the camera. In recent years, with the expansion of the size of star maps, the number of star points in the field of view has increased at a quadratic scale, making it difficult for serial processors to complete the association and extraction tasks within the specified time. It is necessary to design customized computing units in the field to build a fast association and extraction system for space targets. Summary of the invention

[0009] The purpose of the present invention is to overcome the above-mentioned defects and provide a space target rapid association and extraction system and method, which solves the technical problem that the computing power of traditional serial processors cannot meet the computing requirements of multi-frame data processing. The present invention can effectively detect and track space targets and their motion trajectory characteristics, has high computing parallelism and low computing delay, and has good practicality and promotion value.

[0010] In order to achieve the above-mentioned object of the invention, the present invention provides the following technical solutions:

[0011] A space target rapid association and extraction system, comprising: a target point receiving module, a characteristic star map building module, an initial trajectory association module and a suspected / real trajectory expansion module;

[0012] The target point receiving module receives the star point target information and caches the star point target information;

[0013] The characteristic star map construction module receives the star point target information, constructs the characteristic star map according to the star point target information, and outputs the characteristic star map to the initial trajectory association module and the suspected / real trajectory expansion module;

[0014] The initial trajectory association module extracts the star point target information cached by the target point receiving module, and receives the characteristic star map input by the characteristic star map construction module, and associates the suspected trajectories between frames according to the star point target information and the characteristic star map to obtain the initial suspected trajectory, and outputs the initial suspected trajectory to the suspected / true trajectory expansion module;

[0015] The suspected / real trajectory expansion module receives the characteristic star map input by the characteristic star map construction module and the initial suspected trajectory input by the initial trajectory association module, and expands the initial suspected trajectory into a real trajectory according to the characteristic star map.

[0016] Furthermore, the target point receiving module, the characteristic star map building module, the initial trajectory association module and the suspected / real trajectory expansion module are arranged on the FPGA, and the star point target information is inputted through the GTX interface of the FPGA.

[0017] Furthermore, it also includes a target point cache module, a continuous frame feature star map cache module, a suspected trajectory cache module and a real trajectory cache module arranged on the DDR;

[0018] The DDR and FPGA interface is implemented using the MIG interface and address arbitration mapping;

[0019] The target point cache module, the continuous frame characteristic star map cache module, the suspected trajectory cache module and the real trajectory cache module are used to cache the star point target information, the characteristic star map, the suspected trajectory and the real trajectory respectively.

[0020] Furthermore, the method in which the suspected / real trajectory expansion module expands the initial suspected trajectory into a real trajectory according to the characteristic star map includes:

[0021] When a new characteristic star map is received, the potential target points of the characteristic star map are searched with a preset search radius, and the suspected trajectory is extended on the new characteristic star map according to the potential target points to obtain a new suspected trajectory, and whether the extension is valid is judged. If the extension is valid, the confidence of the new suspected trajectory is increased, and if the extension is invalid, the confidence of the new suspected trajectory is reduced;

[0022] Repeat the above steps. When the confidence of the new suspected trajectory reaches the set threshold, update the suspected trajectory to the true trajectory.

[0023] Find the potential target points of the newly received characteristic star map with a preset search radius, expand the real trajectory on the new characteristic star map according to the potential target points, and judge whether the expansion is effective. When the expansion is effective, increase the confidence, and when the expansion is invalid, reduce the confidence;

[0024] When the confidence level is lower than the set threshold, the true trajectory is considered invalid;

[0025] When expanding a suspected trajectory or a true trajectory, the two latest trajectory points of the current trajectory are used to predict the next trajectory point.

[0026] Furthermore, the DDR is divided into areas for ping-pong storage of trajectory data and trajectory information, and FPGA is used to implement a hierarchical storage mechanism for trajectory data and trajectory information;

[0027] The FPGA stores the information page data of suspected trajectories and real trajectories;

[0028] That is, for all traces, the FPGA stores information pages, and the DDR stores the entire trace data;

[0029] The information page data of each track includes the latest two track point data of each track, the number of track points and the storage address of the entire track in DDR.

[0030] Furthermore, the suspected / real trajectory expansion module first loads the information page data of the trajectory each time it expands the trajectory, completes the expansion of the trajectory according to the information page data, and updates the confidence of the trajectory. If the trajectory confidence is reduced to 0 after the update, the trajectory data will no longer be processed (no longer updated, and the storage space of the trajectory will be overwritten if a new trajectory is found later), and the storage space of the trajectory and its information will be overwritten by the newly discovered trajectory. Only the trajectory data with a confidence greater than 0 is loaded and updated and then put back into the DDR. Specifically, when a trajectory is judged invalid, the next new image will not be processed, and the information page will not be loaded. This attempt uses the two latest trajectory points in the past to predict whether there are matching star points in the current image. If there are star points at the predicted position, continue to expand and keep it valid. If there are no star points, reduce the confidence by one. However, if the confidence is greater than 0, the predicted point will still be used to update the trajectory until the confidence is reduced to 0, and the trajectory is "discarded" and considered invalid. The concept of invalidity appears in the prediction-update process, and the trajectory information page that is determined to be invalid will also be discarded and will not be processed again.

[0031] Furthermore, after the suspected / real trajectory expansion module completes the processing of each frame of the feature star map, it switches the spatial address of the trajectory processing cache through a ping-pong operation, and converts the trajectory into a sequential operation of the sequentially increasing address through appending, inserting or removing operations, without updating all trajectory addresses. In the process of suspected / real trajectory expansion, the expansion involves using trajectory information and data to predict the coordinate area where there should be star points in the star map, and then scanning the area in the star map to see if there are star points. If there are, the coordinates of the star points are used to update the latest point of the trajectory and update the confidence. If not, a coordinate is predicted using the two most recent points in the past and appended to the trajectory.

[0032] Furthermore, an on-chip counter is provided on the FPGA, and the on-chip counter is used to number the number of frames of the characteristic star map;

[0033] Every time the suspected / real trajectory expansion module receives a frame of characteristic star map, the on-chip counter is incremented by 1 to complete the switching between the lowest bits 0 and 1 of the characteristic star map number;

[0034] When the lowest bit of the counter number is 1, the suspected / real trajectory expansion module loads the existing trajectory data from the first trajectory information storage space and trajectory data storage space of the ping-pong cache for expansion, and puts the new suspected trajectory obtained by associating the initial trajectory and the trajectory after the real / suspected trajectory is expanded and confirmed into the second trajectory information storage space and trajectory data storage space of the ping-pong cache; and when the lowest bit of the on-chip counter number is 0, the suspected / real trajectory expansion module loads the existing trajectory data from the second trajectory information storage space and trajectory data storage space of the ping-pong cache, and puts the new suspected trajectory obtained by associating the initial trajectory and the trajectory after the real / suspected trajectory is expanded and confirmed into the first trajectory information storage space and trajectory data storage space of the ping-pong cache;

[0035] Append and insert operations are unified as operations on continuously increasing addresses, and remove operations are converted to no processing. The address space is directly overwritten by the new track during the next write.

[0036] Furthermore, both the initial trajectory association module and the suspected / real trajectory expansion module include a pixel parallel processing unit for parallel processing of the pixels of the segmented and loaded image.

[0037] A method for rapid association and extraction of space targets, implemented by using the above-mentioned system for rapid association and extraction of space targets, comprises:

[0038] The initial trajectory association module associates the suspected trajectories between frames according to the star point target information and the characteristic star map to obtain the initial suspected trajectory, and outputs the initial suspected trajectory to the suspected / true trajectory expansion module;

[0039] The suspected / real trajectory expansion module receives the characteristic star map input by the characteristic star map construction module and the initial suspected trajectory input by the initial trajectory association module, and expands the initial suspected trajectory into a real trajectory according to the characteristic star map.

[0040] Furthermore, the initial trajectory association module reads a star point data of the k-th frame feature star map at a time, and traverses and expands to the k+1-th frame feature star map. The traversal operation involves the slice access operation and logical judgment operation of the image data, and the parallel unit is used to realize fast data processing; k≥1;

[0041] Each time a region of a predetermined size is loaded in the k+2th frame image, a valid image slice position is found, and after confirming the pixel positions of the kth, k+1th, and k+2th frames that can be associated, the generated information page of the suspected trajectory and the trajectory points are put into the DDR;

[0042] The suspected / real trajectory expansion module loads the complete suspected / real trajectory data from DDR according to the number of trajectory points and trajectory storage address in the trajectory information page;

[0043] For a single track, before updating the track data, first load the track information, use the latest two track points stored in the information page to predict the potential area of ​​the star point in the next frame of the feature star map, and use the pipeline method to load the feature star map area slice data from DDR in sequence, and process the data. Only when the target star point is found or the confidence is greater than 0 after the track is updated, the track expansion operation is carried out;

[0044] Expand the updated trajectory, insert the pixel position of the valid data of the feature star map into the trajectory data of the old trajectory, send it to DDR, and update the trajectory information;

[0045] Send the updated trajectory information to the DDR buffer.

[0046] The space target fast association and extraction system of the present invention belongs to the field of high-performance computing technology in the field of computer technology. Specifically, it relates to space target detection and tracking technology, especially based on hardware logic units, building a customized computing system to achieve fast space target detection and tracking technology. Compared with the commonly used target detection and tracking methods, the targets in the space image have the characteristics of more noise, high similarity between the target and the noise, and fewer samples available for parameter training. It is difficult to rely on traditional image processing methods to effectively obtain the target point and its motion trajectory from a single frame image. It is necessary to rely on the fusion processing and comparison of multi-frame data to achieve effective detection and tracking of the target. The fusion processing of multi-frame data involves complex external memory access and data indexing operations. Under the trend of increasing star points in the field of view, the computing power of traditional serial processors can no longer meet the computing requirements of multi-frame data processing. In addition, the update interval of the space image is fixed. If the computing task cannot be completed within the frame interval, it will cause congestion in the entire computing process and affect the execution of the entire system function. Therefore, it is necessary to design a computing system suitable for space target association and extraction in a targeted manner, improve the parallelism of the calculation, reduce the computing delay, and achieve stable operation of the entire computing process. Therefore, relevant technical fields include computing architecture design, high-performance computing, image processing technology, timing optimization and other aspects.

[0047] Compared with the prior art, the present invention has at least one of the following beneficial effects:

[0048] (1) This invention is the first to implement the traditional serial processing target association and extraction algorithm using hardware logic, with complete initial trajectory association, suspected trajectory expansion, real trajectory expansion and other functions;

[0049] (2) The present invention adopts a computational parallelization design, which effectively reduces the computational delay of the entire target association and extraction, improves the system throughput, and realizes rapid data processing;

[0050] (3) The present invention effectively reduces the number of memory access operations during track appending, insertion and deletion through DDR ping-pong processing operations, thereby reducing the memory access complexity;

[0051] (4) The layered space target trajectory and information storage strategy of the present invention effectively reduces unnecessary storage access operations and improves data utilization efficiency;

[0052] (5) In the present invention, the calculation units of initial trajectory association, suspected trajectory extension and real trajectory extension are all implemented in a modular way, which simplifies the implementation path and resource usage, facilitates subsequent maintenance and updating, and can be tailored and reused according to resource usage. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] Figure 1 This is a schematic diagram of the initial trajectory association of the present invention;

[0054] Figure 2 It is a schematic diagram of trajectory growth and confirmation of the present invention;

[0055] Figure 3 This is a schematic diagram of the composition of the space target rapid association and extraction system of the present invention;

[0056] Figure 4 A schematic diagram of the layered trajectory data cache of the present invention;

[0057] Figure 5 This is a schematic diagram of ping-pong cache switching in the present invention;

[0058] Figure 6 Schematic diagram of the deployment architecture of the rapid target association and extraction system of the present invention. DETAILED DESCRIPTION

[0059] The following detailed description of the present invention will make the features and advantages of the present invention more clear and explicit.

[0060] The word "exemplary" is used exclusively herein to mean "serving as an example, embodiment, or illustration." Any embodiment described herein as "exemplary" is not necessarily to be construed as preferred or advantageous over other embodiments. Although various aspects of the embodiments are shown in the drawings, the drawings are not necessarily drawn to scale unless otherwise noted.

[0061] The present invention proposes and implements a space target rapid association and extraction system from the perspectives of hardware implementation, calculation optimization and system design.

[0062] like Figure 1 and Figure 2 ,Space target association and extraction includes three parts: initial trajectory association, suspected trajectory expansion and real trajectory expansion.,The initial trajectory association traverses the star points in the image from three consecutive frames,,starting from the first frame, searches for the target of the next frame with a set constraint radius,,connects the adjacent star point data in the three frames in series, and forms the initial suspected trajectory.

[0063] After the suspected trajectory is constructed, each time new graphic star point data is received, new potential target points are continued to be searched with the search radius, the suspected trajectory is extended on the new feature star map frame, and the trajectory confidence is increased or decreased according to whether the extension is effective.

[0064] When the confidence of the suspected trajectory reaches the set threshold, the suspected trajectory is updated to the real trajectory. The real trajectory continues to perform the expansion task of each frame and updates the trajectory confidence information. When the trajectory confidence is lower than the set threshold, the trajectory is considered invalid and no subsequent expansion and update are performed.

[0065] The above calculation steps need to execute a large number of traversal and local image search steps, frequently access star point and image data, and perform a large number of append and insert operations on the trajectory. With the continuous improvement of spatial information perception requirements, the calculation and memory access architecture and the number of computing units of traditional serial processors can no longer meet the actual data update requirements. Figure 3 shown.

[0066] The target fast association and extraction system is built through the hardware logic unit. First, the star point target information (including star points and target points) is received through the high-speed GTX interface. DDR is used as DDR, and the MIG interface and address arbitration mapping are used to cache the runtime data and trajectory information. The parallel processing unit uses the hardware logic unit to build a parallel computing and processing array to achieve a peak computing power of 100GOPS (@200MHZ), which mainly includes the initial trajectory association part to realize the suspected trajectory construction, the suspected / real trajectory expansion part to expand and update the suspected trajectory and the real trajectory, and upgrade the suspected trajectory that reaches the set threshold to the real trajectory.

[0067] like Figure 4, based on the high computing power achieved by stacking multi-way parallel computing paths, each trajectory expansion only requires the first two trajectory point data of the current trajectory to predict the next trajectory point. Each time the trajectory is processed, loading all trajectories requires an additional external storage access process. Therefore, a layered trajectory data cache architecture is designed to construct the information of each trajectory according to the data organization form in the figure below. The information of each trajectory contains the latest two trajectory point data, the trajectory point reading, and the storage address of the entire trajectory data. Each time the trajectory is expanded, the trajectory information page data is loaded first to complete the confirmation of the trajectory expansion. If the trajectory confidence is reduced to 0 after the update, the trajectory data will not be processed directly; only the valid trajectory data will be loaded and updated in the next step.

[0068] In addition, after each frame of image processing is completed, the spatial address of the trajectory processing cache is switched through a ping-pong operation, and the trajectory append / insert / remove operations are converted into sequential and orderly operations on sequentially increasing addresses. There is no need to update all trajectory addresses, which effectively saves processing time.

[0069] like Figure 5 , an on-chip counter is used to count the image frames. Each time the star point data of a frame of image is received, the on-chip counter is increased by 1 to complete the switching of the lowest bit '0' and '1'. When the lowest bit of the image number is '1', the existing trajectory data is loaded from the first trajectory information storage space and trajectory data storage space for the real / suspected trajectory expansion and confirmation calculation process to process, and the new suspected trajectory obtained by associating the initial trajectory and the trajectory after the real / suspected trajectory is expanded and confirmed are placed in the second trajectory information storage space and trajectory data storage space; and when the lowest bit of the counter is '0', the existing trajectory data is loaded from the second trajectory information storage space and trajectory data storage space for the real / suspected trajectory expansion calculation process to process, and the new suspected trajectory obtained by associating the initial trajectory and the trajectory after the real / suspected trajectory is expanded and confirmed are placed in the second trajectory information storage space and trajectory data storage space. The append and insert operations are unified to continuously increment the continuous address, and the elimination operation is converted to no processing, and the address space is directly overwritten by the new trajectory when writing next time.

[0070] In summary, the space target rapid association and extraction system of the present invention is implemented by logic unit hardware. First, the star point data in the frame image is received to construct a characteristic star map, and then the three calculation steps of real trajectory extension, suspected trajectory extension, and initial trajectory association are sequentially executed to extract the moving space target from a large amount of star point data and construct its motion trajectory characteristics. The parallel and pipeline strategies of data access and calculation are used to ensure the high speed and low latency of the entire calculation process. The hierarchical trajectory storage strategy is adopted to reduce the access requirements for trajectory processing and construction to external storage, and the ping-pong strategy is designed for cached data to achieve efficient data appending and insertion. The space target rapid association and extraction system of the present invention can be widely deployed in embedded boards and stand-alone machines, and applied to the rapid detection and tracking of space targets. In addition, the technical solution of the present invention has high computational parallelism and low computational latency, as well as low cost and small hardware resource occupancy, and can effectively detect and track space targets and their motion trajectory characteristics, and has good practicality and promotion value.

[0071] Example:

[0072] The hardware architecture of the present invention is as follows Figure 6 As shown in the figure, the pre-processing module completes the star point detection and transmits the star point data to the target association and extraction system through the high-speed bus. The target association and extraction system completes the extraction of space targets and the construction of motion trajectories. During the operation, a large amount of cache data is stored in DDRDDR3. The final result is transmitted to the DSP / CPU or telemetry terminal for subsequent data calculation and processing.

[0073] The joint simulation of the overall function of trajectory association was completed on the PC. The test function was correct and the simulation results were consistent with the serial processor operation results.

[0074] (1) Initial trajectory association

[0075] When the trajectory association starts, the star point data is loaded first, and the star point data in the DDR is loaded into the on-chip cache in batches for processing by the association module (i.e., the initial trajectory association module and the suspected / real trajectory expansion module).

[0076] The data of a star point in the characteristic star map of the kth frame is read once, and the data is traversed and extended to the k+1th frame image. The traversal operation involves the slice access operation and logical judgment operation of the image data, and the parallel unit is used to realize fast data processing.

[0077] Each time the 7×7 area is loaded in the k+2th frame image, a valid image slice position is found, and after confirming the pixel positions of the kth, k+1th and k+2th frames that can be associated, the trajectory information and trajectory points are put into the DDR.

[0078] (2) Expansion of suspected / real trajectories

[0079] Load suspected trajectory information data from the trajectory information storage space of DDR. When there is a lot of trajectory information, adopt a batch loading strategy, load only one batch of information data each time, put it into the on-chip storage, and then use a pipeline method to process the data of each trajectory information in turn.

[0080] From the current trajectory information data, extract the latest two trajectory points to predict the potential area of ​​the next frame, and then use the pipeline method to load the image area slice data from the feature star map cache in turn to process the data. Use the potential area to find valid star points and their locations to expand the trajectory with a confidence greater than 0.

[0081] Expand the updated trajectory, append the new trajectory points to the trajectory data of the old trajectory, send it to the external trajectory data cache, and update the trajectory information.

[0082] Send the updated trajectory information to the trajectory information cache.

[0083] The actual operation test verification of the function was realized on the FPGA board. After testing, in the case of 8000 target points per frame in a typical scenario, the FPGA processing time at 100M main frequency is 833ms, and the processing time at 200M main frequency is less than 500ms, which meets the actual use requirements. The module is logically synthesized on the Xilinx platform, and the resource consumption is shown in Table 1. Compared with DSP, the performance is improved by 84 times at 100M main frequency and more than 160 times at 200M main frequency. It has small resource occupation and high operation efficiency, which is easy to integrate and deploy in a variety of embedded boards and stand-alone platforms.

[0084] Table 1 Resource consumption statistics

[0085]

[0086] The present invention has been described in detail above in conjunction with specific implementations and exemplary examples, but these descriptions cannot be understood as limiting the present invention. Those skilled in the art understand that, without departing from the spirit and scope of the present invention, a variety of equivalent substitutions, modifications or improvements may be made to the technical solution of the present invention and its implementation methods, all of which fall within the scope of the present invention. The scope of protection of the present invention shall be subject to the attached claims.

[0087] The contents not described in detail in the specification of the present invention belong to the common knowledge of those skilled in the art.

Claims

1. A space target rapid association and extraction system, characterized in that: include: Target point receiving module, characteristic star map building module, initial trajectory association module and suspected / real trajectory expansion module; The target point receiving module receives the star point target information and caches the star point target information; The characteristic star map construction module receives the star point target information, constructs the characteristic star map according to the star point target information, and outputs the characteristic star map to the initial trajectory association module and the suspected / real trajectory expansion module; The initial trajectory association module extracts the star point target information cached by the target point receiving module, and receives the characteristic star map input by the characteristic star map construction module, and associates the suspected trajectories between frames according to the star point target information and the characteristic star map to obtain the initial suspected trajectory, and outputs the initial suspected trajectory to the suspected / true trajectory expansion module; The suspected / real trajectory expansion module receives the characteristic star map input by the characteristic star map construction module and the initial suspected trajectory input by the initial trajectory association module, and expands the initial suspected trajectory into a real trajectory according to the characteristic star map.

2. A space target rapid association and extraction system according to claim 1, characterized in that: The target point receiving module, characteristic star map building module, initial trajectory association module and suspected / real trajectory expansion module are set on the FPGA, and the star point target information is input through the GTX interface of the FPGA; Both the initial trajectory association module and the suspected / real trajectory expansion module include a pixel parallel processing unit for parallel processing of the pixels of the segmented and loaded images.

3. A space target rapid association and extraction system according to claim 2, characterized in that: It also includes a target point cache module, a continuous frame feature star map cache module, a suspected trajectory cache module and a real trajectory cache module arranged on the DDR; The DDR and FPGA interface is implemented using the MIG interface and address arbitration mapping; The target point cache module, the continuous frame characteristic star map cache module, the suspected trajectory cache module and the real trajectory cache module are used to cache the star point target information, the characteristic star map, the suspected trajectory and the real trajectory respectively.

4. A space target rapid association and extraction system according to claim 3, characterized in that: The method of the suspected / real trajectory expansion module expanding the initial suspected trajectory into the real trajectory according to the characteristic star map includes: When a new characteristic star map is received, the potential target points of the characteristic star map are searched with a preset search radius, and the suspected trajectory is extended on the new characteristic star map according to the potential target points to obtain a new suspected trajectory, and whether the extension is valid is judged. If the extension is valid, the confidence of the new suspected trajectory is increased, and if the extension is invalid, the confidence of the new suspected trajectory is reduced; Repeat the above steps. When the confidence of the new suspected trajectory reaches the set threshold, update the suspected trajectory to the true trajectory. Find the potential target points of the newly received characteristic star map with a preset search radius, expand the real trajectory on the new characteristic star map according to the potential target points, and judge whether the expansion is effective. When the expansion is effective, increase the confidence, and when the expansion is invalid, reduce the confidence; When the confidence level is lower than the set threshold, the true trajectory is considered invalid; When expanding a suspected trajectory or a true trajectory, the two latest trajectory points of the current trajectory are used to predict the next trajectory point.

5. A space target rapid association and extraction system according to claim 4, characterized in that: The FPGA stores the information page data of suspected trajectories and real trajectories; The information page data of each track includes the latest two track point data of each track, the number of track points and the storage address of the entire track in DDR.

6. A space target rapid association and extraction system according to claim 5, characterized in that: Each time the suspected / real trajectory expansion module performs trajectory expansion, it first loads the information page data of the trajectory, completes the expansion of the trajectory based on the information page data, and updates the confidence of the trajectory. If the trajectory confidence drops to 0 after the update, the trajectory data will no longer be processed, and the storage space of the trajectory and its information will be overwritten by the newly discovered trajectory. Only the trajectory data with a confidence greater than 0 will be loaded and updated and then put back into the DDR.

7. A space target rapid association and extraction system according to claim 6, characterized in that: After processing each frame of the feature star map, the suspected / real trajectory expansion module switches the spatial address of the trajectory processing cache through a ping-pong operation, and converts the trajectory into sequential operations on sequentially increasing addresses through append, insert or delete operations, without updating all trajectory addresses.

8. A space target rapid association and extraction system according to claim 7, characterized in that: An on-chip counter is provided on the FPGA, and the on-chip counter is used to number the number of frames of the characteristic star map; DDR divides the ping-pong buffer and the area for storing trajectory information; Every time the suspected / real trajectory expansion module receives a frame of characteristic star map, the on-chip counter is incremented by 1 to complete the switching between the lowest bits 0 and 1 of the characteristic star map number; When the lowest bit of the counter number is 1, the suspected / real trajectory expansion module loads the existing trajectory data from the first trajectory information storage space and trajectory data storage space of the ping-pong cache for expansion, and puts the new suspected trajectory obtained by associating the initial trajectory and the trajectory after the real / suspected trajectory is expanded and confirmed into the second trajectory information storage space and trajectory data storage space of the ping-pong cache; and when the lowest bit of the on-chip counter number is 0, the suspected / real trajectory expansion module loads the existing trajectory data from the second trajectory information storage space and trajectory data storage space of the ping-pong cache, and puts the new suspected trajectory obtained by associating the initial trajectory and the trajectory after the real / suspected trajectory is expanded and confirmed into the first trajectory information storage space and trajectory data storage space of the ping-pong cache; Append and insert operations are unified as operations on continuously increasing addresses, and remove operations are converted to no processing. The address space is directly overwritten by the new track during the next write.

9. A method for rapid association and extraction of space targets, characterized in that: The method is implemented by using a space target rapid association and extraction system as described in any one of claims 1 to 8, comprising: The initial trajectory association module associates the suspected trajectories between frames according to the star point target information and the characteristic star map to obtain the initial suspected trajectory, and outputs the initial suspected trajectory to the suspected / true trajectory expansion module; The suspected / real trajectory expansion module receives the characteristic star map input by the characteristic star map construction module and the initial suspected trajectory input by the initial trajectory association module, and expands the initial suspected trajectory into a real trajectory according to the characteristic star map.

10. A method for rapid association and extraction of space targets according to claim 9, characterized in that: The initial trajectory association module reads a star point data of the k-th frame feature star map at a time, and traverses and expands to the k+1-th frame feature star map. The traversal operation involves the slice access operation and logical judgment operation of the image data, and the parallel unit is used to realize fast data processing; k≥1; Each time a region of a predetermined size is loaded in the k+2th frame image, a valid image slice position is found, and after confirming the pixel positions of the kth, k+1th, and k+2th frames that can be associated, the generated information page of the suspected trajectory and the trajectory points are put into the DDR; The suspected / real trajectory expansion module loads the complete suspected / real trajectory data from DDR according to the number of trajectory points and trajectory storage address in the trajectory information page; For a single track, before updating the track data, first load the track information, use the latest two track points stored in the information page to predict the potential area of ​​the star point in the next frame of the feature star map, and use the pipeline method to load the feature star map area slice data from DDR in sequence, and process the data. Only when the target star point is found or the confidence is greater than 0 after the track is updated, the track expansion operation is carried out; Expand the updated trajectory, insert the pixel position where the valid data of the feature star map is located into the trajectory data of the old trajectory, send it to DDR, and update the trajectory information; Send the updated trajectory information to the DDR buffer.

Images