Autonomous target tracking method for medium and high error

Abstract

The invention discloses an out-of-satellite information correction-based target autonomous tracking method for medium and high orbit remote sensing satellites, and belongs to the technical field of satellite task execution. According to the method, when a satellite fails to correctly detect a target concerned by a user, target approximate position information from an off-satellite information source is introduced to correct a task, and the target concerned by the user can be detected at a higher probability by detecting an area near the approximate position again and adjusting a confidence threshold value of a detection algorithm; the follow-up tracking task can be normally executed, and the success rate of the task is effectively improved.

Description

Technical Field

[0001] The invention relates to a target autonomous tracking method for a medium- and high-orbit remote sensing satellite based on extraterrestrial information correction, belonging to the technical field of satellite mission execution. Background Art

[0002] Compared with low-orbit satellites, medium- and high-orbit remote sensing satellites have the advantage of staying in a fixed area for a longer time, so they are more suitable for continuous tracking of targets. However, the imaging resolution of medium- and high-orbit remote sensing satellites is weaker than that of low-orbit satellites, resulting in a decrease in the detection capability of targets, which is not conducive to the correct detection of tracking targets.

[0003] The satellite's autonomous tracking capability for targets is based on the onboard data processing function. That is, the satellite processes the image, finds the target to be tracked from the image and calculates its trajectory, and autonomously decides to adjust the satellite imaging direction at the appropriate time to keep the target in the imaging field of view, thereby achieving autonomous tracking of the target. However, due to the insufficient satellite detection capability, onboard data processing capability, and imaging background interference such as clouds, fog, and waves, the results of onboard data detection often contain multiple suspected targets, which may not include the target of concern to the user.

[0004] There are two existing methods for satellite autonomous tracking and target selection: one is that the satellite autonomously selects a target for tracking. The disadvantage is that the target autonomously selected by the satellite is often not the target that the user really cares about; the other is that the satellite first transmits all suspected targets detected to ground users, and the ground users select the targets they care about and feed them back to the satellite for subsequent satellite autonomous tracking. The disadvantage is that if all suspected targets detected by the satellite are not the targets that the user cares about, the mission will fail, wasting the satellite's precious mission time. If the user wants to continue the tracking mission, it is necessary to re-coordinate the satellite and ground units to initiate a new mission, which takes a lot of time and is still difficult to avoid encountering the same problem. The shortcomings of the above existing methods have greatly weakened the practicality of the current satellite autonomous tracking function. Summary of the invention

[0005] The technical problem solved by the present invention is: to overcome the shortcomings of the prior art, propose an autonomous target tracking method for medium and high orbit remote sensing satellites based on extraterrestrial information correction, correct the onboard data processing results and satellite tracking task execution, and realize accurate and continuous tracking of targets by medium and high orbit remote sensing satellites.

[0006] The technical solution of the present invention is:

[0007] A method for autonomous target tracking of a medium-orbit remote sensing satellite based on extraterrestrial information correction, comprising:

[0008] Step 1: The satellite performs tasks according to user requirements, continuously images a specified area, and sends the generated image sequence to the ground and on-board data processing modules;

[0009] Step 2: The on-board data processing module uses a target detection algorithm to process the image sequence, sets the confidence threshold of the target detection algorithm, detects all suspected targets in the images that are above the confidence threshold, and calculates the position information of each suspected target;

[0010] Step 3: The on-board data processing module continuously processes multiple scenes of images in the image sequence, calculates the motion trajectory information of each suspected target, numbers each suspected target, and then transmits the suspected target information to the ground;

[0011] Step 4: The ground user confirms the suspected target information output by the satellite. If the multiple suspected targets output by the satellite include the target concerned by the user, then proceed to Step 5; if the multiple suspected targets output by the satellite do not include the target concerned by the user, then proceed to Step 6;

[0012] Step 5: The ground user selects the concerned target from the satellite output results and uploads the target information containing the target number to the satellite; the satellite continuously generates the trajectory and position information of this target by continuously processing the image imaging sequence, realizing continuous tracking of the target until the satellite ends the tracking task under the control of the ground user;

[0013] Step 6: The ground user conducts comprehensive analysis and judgment using off-satellite information to obtain the approximate position of the target concerned by the user and uploads it to the satellite; the satellite forms a preferred area in image processing with the approximate position information as the center;

[0014] Step 7: The on-board data processing module retrieves all suspected targets in the images again according to Step 2; among them, for areas outside the preferred area, the screening of suspected targets is still carried out according to the currently set threshold; for the preferred area, the currently set confidence threshold is reduced; if suspected target information is detected in this round, calculate the position information of each suspected target and transmit it to the ground again, and return to Step 4; if no suspected target information is detected in this round and the confidence threshold corresponding to the preferred area is greater than 0, continue to reduce the currently set confidence threshold of the preferred area and retrieve all suspected targets in the images again; if no suspected target information is detected in this round and the confidence threshold corresponding to the preferred area is 0, then the satellite ends the tracking task.

[0015] Furthermore, the off-satellite information includes but is not limited to the interpretation results obtained by manual or machine calculation of the satellite images on the ground, the imaging and calculation results from other satellites, the detection results of ground-based / space-based radars, and the signals of the AIS automatic identification system.

[0016] Further, in step 5, during the continuous tracking of the target by the satellite, the trajectory and position information of the target are continuously generated, and the distance of the target from the edge of the satellite imaging field of view is calculated.

[0017] If the distance of the target from the edge of the satellite imaging field of view is less than the set pointing adjustment threshold Q, the satellite imaging pointing is adjusted to move the center of the imaging field of view to the target position, so that the target returns to the central position area of the imaging area.

[0018] Further, the pointing adjustment threshold Q is expressed in terms of the number of pixels, and the Q value is preferably 5% - 30% of the imaging field of view.

[0019] Further, in step 3, the suspected target information includes but is not limited to the target number, target type, target position and trajectory, target confidence, and target detection time.

[0020] Further, in step 2, the target detection algorithms include but are not limited to Yolo, SSD, RetinaNet; the confidence of the target detection algorithm is a value attached to each detection result in the target detection task by the algorithm, representing the credibility of the result, and the range of the confidence is from 0 to 1, where 0 means the algorithm does not believe the result at all, and 1 means the algorithm is sure that the result is correct.

[0021] Further, in step 2, the initial value of the confidence threshold of the target detection algorithm is preferably 0.3 - 0.7.

[0022] Further, in step 6, the preferred area setting value is preferably 2% - 20% of the imaging field of view.

[0023] Further, in step 1, the task settings include the imaging time, imaging interval, imaging pointing position, and task duration; among them, the imaging interval is preferably less than 10 minutes.

[0024] Further, in step 5, the satellite adopts a target tracking algorithm to complete the association of the same target in multiple scene sequence images by continuously processing the image imaging sequence, form the target motion trajectory, and continuously generate the trajectory and position information of the target.

[0025] The advantages of the present invention compared with the prior art are as follows:

[0026] (1) Without changing the satellite detection ability and on - board target processing ability, the present invention makes full use of multi - source information available outside the satellite, reduces the range of the target area, improves the detection success probability of the target concerned by the user by the satellite, and improves the problems that in the current autonomous tracking process of medium - and high - orbit remote sensing satellites, there are many detected targets but inaccurate ones, and the target tracking is invalid or fails.

[0027] (2) In the present invention, when the satellite fails to autonomously detect a target, the problem that the mission directly fails due to the lack of adjustment ability in the traditional method is avoided, ensuring the normal operation of the entire autonomous tracking process.

[0028] (3) In the traditional target detection, positioning, selection, and tracking process of the present invention, a correction mechanism based on extra-satellite information is introduced, making full use of the multi-source external information available outside the satellite in current satellite applications, taking into account the autonomy of on-board processing and planning and the controllability of ground user correction, effectively improving the accuracy of detecting targets in on-board data processing, and enhancing the success rate, flexibility, and practicality of the satellite autonomous tracking mission. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] By reading the following detailed description of the preferred embodiments, various other advantages and benefits will become clear to those of ordinary skill in the art. The drawings are only for the purpose of illustrating the preferred embodiments and are not to be considered as limiting the present invention. Moreover, throughout the drawings, the same reference numerals are used to represent the same components. In the drawings:

[0030] Figure 1 It is a flowchart of the method for autonomous target tracking of a geostationary remote sensing satellite based on extra-satellite information correction in an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0031] The exemplary embodiments of the present disclosure will be described in more detail below with reference to the drawings. Although the exemplary embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure can be implemented in various forms and should not be limited by the embodiments set forth herein. On the contrary, these embodiments are provided so that the present disclosure can be more thoroughly understood and the scope of the present disclosure can be fully conveyed to those skilled in the art.

[0032] According to GB / T 29079-2012 "Spacecraft Orbit Classification and Common Parameter Symbols", an orbit with an altitude lower than 2000 km is defined as a low orbit, an orbit with an altitude above 30000 km is defined as a high orbit, and a medium orbit is defined as an orbit altitude between the low orbit and the high orbit.

[0033] The present invention proposes a method for autonomous target tracking of medium and high orbit remote sensing satellites based on correction by extra-satellite information, which is applicable to remote sensing satellites with an orbital altitude higher than 2000 km. This method requires an on-board data processing module, which is a typical on-board module configured for satellites with certain requirements for the timeliness of image processing. It is generally configured with high computing power chips such as high-performance FPGAs, DSPs, and GPUs, and can realize the on-orbit rapid processing of the imaging of payloads such as satellite cameras, and complete typical tasks such as cloud discrimination, radiation correction, target detection, target positioning, target tracking, and region extraction. The results generated by its processing are sent to users in need through the satellite's external transmission channels such as the downlink data transmission channel and the downlink broadcast channel to achieve high-timeliness information guarantee services for users.

[0034] This method is as Figure 1 shown and includes the following steps:

[0035] Step 1, the satellite executes tasks according to user requirements and continuously images a specified area. The imaging methods include but are not limited to optics (visible light, infrared, hyperspectral, etc.), SAR. The generated image sequence is sent to the ground, and at the same time, the image sequence is sent to the on-board data processing module.

[0036] The task settings should include imaging time, imaging interval, imaging pointing position, task duration, etc. Among them, the imaging interval setting should comprehensively consider the processing capacity of the on-board data processing module and the need to complete scene-to-scene association of the target, and is preferably less than 10 minutes.

[0037] Step 2, the on-board data processing module processes the image and uses the target detection algorithm to complete the detection of suspected targets in the image. Among them, the target detection algorithms include but are not limited to various algorithms suitable for target detection such as Yolo, SSD, RetinaNet, etc. Confidence is a value attached to each detection result by the algorithm in the target detection task to represent the credibility of the result. The range of confidence is from 0 to 1. 0 means that the algorithm completely does not believe the result, and 1 means that the algorithm is very sure that the result is correct.

[0038] In each target processing task, by setting different confidence thresholds, the effect of adjusting the output results can be achieved. That is, when the confidence threshold is set relatively high, targets with relatively low confidence can be filtered out to prevent too many false targets and incorrect targets from being output, and at the same time, it can reduce the amount of data output by the satellite and relieve the satellite-ground transmission pressure; when the confidence threshold is set relatively low, it can help some weak and uncertain targets to be retained in the output results. In step 8, according to user feedback, the confidence threshold can be flexibly adjusted during a task to improve the detection success rate of the targets concerned by users.

[0039] In this embodiment, the confidence threshold of the algorithm is initially set to C1. The value range of C1 is between 0 and 1, and it needs to be set by comprehensively considering factors such as the false alarm rate requirement, satellite processing and transmission capacity, and effective detection of targets. It is preferably set between 0.3 and 0.7. The algorithm only outputs suspected targets with a confidence higher than C1 to control the total number of output suspected targets.

[0040] After that, for the output suspected targets, the position information of each target is calculated based on the satellite orbit attitude data and the rigorous imaging model.

[0041] Step 3: The on-board data processing module continuously processes multiple images in the image sequence. Using the target tracking algorithm, it completes the matching and association of each suspected target in multiple images, calculates the motion trajectory information of the suspected target (including the position information of the target in the imaging sequence), numbers each suspected target, and transmits the suspected target information to the ground user. The suspected target information includes but is not limited to the target number (in a single mission, the number of each target needs to be unique), target type (such as different types of ships, etc.), target position and trajectory (expressed in the form of longitude, latitude, altitude, or the CGCS2000 coordinate system, etc.), target confidence, target detection time, and other information.

[0042] Among them, the target tracking algorithm includes but is not limited to traditional algorithms such as Kalman filtering and particle filtering, or deep learning algorithms such as DLT and FCNT, which are used to complete the association of the same target in multiple image sequences, form the target motion trajectory, and are used for subsequent target tracking by the satellite.

[0043] Step 4: The ground user confirms the suspected target information output by the satellite. If the multiple suspected targets output by the satellite include the targets concerned by the user, then step 5 is executed; if the multiple suspected targets output by the satellite do not include the targets concerned by the user, then step 6 is executed.

[0044] Step 5: The user selects the concerned targets from the satellite output results and uploads them to the satellite. The uploaded target information must include the target number. The satellite continuously processes the satellite imaging sequence, continuously generates the trajectory and position information of the target corresponding to the number, and calculates the distance between the target and the edge of the satellite imaging field of view. The distance between the target and the edge of the imaging field of view can be statistically represented by the number of pixels between the target and the edge of the field of view in the image.

[0045] As the target moves and time passes, the target may gradually approach the edge of the satellite imaging field of view, or even leave the field of view. To ensure continuous observation and tracking of the target, it is necessary to adjust the satellite's imaging pointing direction before the target leaves the field of view, and adjust the center of the imaging area back to the vicinity of the target position. Therefore, a pointing adjustment threshold Q is defined. The Q value is expressed in the number of pixels and can be set according to application requirements. A large Q setting value may cause the satellite to frequently adjust its pointing direction, thereby affecting the stable imaging of the target; when the setting value is small, there is a risk of the target leaving the field of view. The setting value of Q is preferably 5% to 30% of the imaging field of view.

[0046] When the distance between the target that the user is concerned about and the edge of the satellite imaging field of view is less than the pointing adjustment threshold Q, the satellite imaging pointing is adjusted to adjust the center of the imaging field of view to the target position, so that the target returns to the vicinity of the center of the imaging area. Repeat step 5 to achieve continuous tracking of the target until the task is ended under user control.

[0047] Step 6: If the suspected target group does not contain the target of interest to the user, the user can use the extra-satellite information to make a comprehensive analysis and judgment, obtain the approximate location of the target of interest to the user, and upload it to the satellite. The satellite forms a preferred area A in the image processing with the approximate location information as the center.

[0048] Extra-satellite information refers to information other than the information obtained by the satellite through its own calculations, including but not limited to the interpretation results obtained by manual or machine calculations of the satellite image on the ground, imaging and calculation results from other satellites, ground-based / space-based radar detection results, AIS automatic identification system signals, etc. For example, professional readers finally confirmed the faint target that was missed by the on-board algorithm through a long period of careful interpretation of the continuously transmitted images. Or, the current probabilistic position of the target is calculated by using the target image temporarily taken by a low-orbit high-resolution satellite when it flew over the target earlier, combined with the manual estimation of the target's speed and heading. This is because the low-orbit satellite has a low orbit and high resolution, so it is easy to find the target, but due to its orbital characteristics, the relative speed with the earth is fast, and the target cannot be continuously monitored. Therefore, it is necessary to calculate the current approximate position of the target based on the location of the target at the time of earlier imaging, combined with the target's moving speed and time.

[0049] The range of the preferred area A can be set. Considering the errors in the extraterrestrial information and the movement of the target from the time the extraterrestrial information is generated to the time it is transmitted to the satellite, the range of A should be set moderately. If A is set too small, the current position of the target may not be covered due to the movement of the target; if the range of A is too large, it will not be able to focus on the key area and exclude interference from irrelevant areas. The setting value of A is preferably 2% to 20% of the imaging field of view.

[0050] Step 7: Perform target detection and positioning processing in the same way as in Step 2. The difference from Step 2 is that for the preferred region A, the algorithm adjusts the confidence threshold to C2 (0≥C2<C1) to relax the detection conditions for suspected targets within the preferred region and increase the detection probability of weak targets that were not detected previously. For the part outside the preferred region, the suspected target screening is still carried out according to the C1 threshold. If suspected target information is detected in this round, calculate the position information of each suspected target, and transmit it to the ground again, then return to Step 4; if no suspected target information is detected in this round and the confidence threshold corresponding to the preferred region is greater than 0, continue to decrease the currently set confidence threshold of the preferred region and retrieve all suspected targets in the image again; if no suspected target information is detected in this round and the confidence threshold corresponding to the preferred region is 0, the satellite ends the tracking task.

[0051] Each time Step 7 is iterated, the set value of the confidence threshold is further decreased. Define the confidence threshold each time as C n (n is an integer greater than or equal to 1), where C1 is the threshold set for the first detection process in Step 2, C2 is the threshold set for the first detection process based on ground correction in Step 7, and C n is the threshold set for the (n - 1)-th detection process based on ground correction. It should satisfy:

[0052]

[0053] Among them, C last represents the confidence threshold set for the last detection attempt on the target concerned by the user according to the tolerable number of detection failures by the user. In this attempt, set the confidence threshold C last to 0, that is, output all targets that can be detected by the algorithm. If the target concerned by the user still cannot be effectively detected when the confidence C last is set to 0, then this task does not meet the conditions for successful execution and the task is terminated.

[0054] In Step 8 and Step 10, the adjustment strategy of the confidence threshold for the detection of the preferred region is determined according to the tolerable number of detection failures by the user and the initial threshold C1. Specifically, during the target tracking process, when the tracked target is lost or changed due to various internal and external factors of the satellite, based on the above method, the ground user can also provide correction information to adjust the current task without terminating the current task.

[0055] In addition to the above-described process, during the target tracking process, when the tracked target is lost or the user's tracking requirement is changed due to various internal and external factors of the satellite, based on the above method, the ground user can also provide correction information to adjust the current task to match the user's requirement without terminating the current task.

[0056] When the satellite fails to correctly detect the target concerned by the user, the present invention introduces the approximate position information of the target from an extra-satellite information source to correct the task. By re-detecting the area near the approximate position and adjusting the confidence threshold of the detection algorithm, the target concerned by the user can be detected with a higher probability, so as to support the normal execution of the subsequent tracking task and effectively improve the success rate of the task.

[0057] The above-described embodiments are only relatively preferred specific embodiments of the present invention. The ordinary changes and substitutions made by those skilled in the art within the scope of the technical solution of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for autonomous target tracking of a medium and high orbit remote sensing satellite based on correction of external information, characterized in that Including: Step 1: The satellite executes tasks according to user requirements, continuously images a specified area, and sends the generated image sequence to the ground and on-board data processing modules; Step 2: The on-board data processing module uses a target detection algorithm to process the image sequence, sets the confidence threshold of the target detection algorithm, detects all suspected targets in the image that are higher than the confidence threshold, and calculates the position information of each suspected target; Step 3: The on-board data processing module continuously processes multiple scenes of images in the image sequence, calculates the motion trajectory information of each suspected target, numbers each suspected target, and then transmits the suspected target information to the ground; Step 4: The ground user confirms the suspected target information output by the satellite. If the multiple suspected targets output by the satellite include the target of interest to the user, then execute Step 5; if the multiple suspected targets output by the satellite do not include the target of concern to the user, then execute Step 6; Step 5: The ground user selects the target of interest from the satellite output results and uploads the target information including the target number to the satellite; the satellite continuously generates the trajectory and position information of the target by continuously processing the image imaging sequence, and realizes continuous tracking of the target until the satellite ends the tracking task under the control of the ground user; Step 6: The ground user conducts comprehensive analysis and judgment using out-of-satellite information to obtain the approximate position of the target of interest to the user, and uploads it to the satellite; the satellite forms a preferred area in image processing with the approximate position information as the center; Step 7: The on-board data processing module retrieves all suspected targets in the image again according to Step 2; among them, for the area outside the preferred area, the suspected targets are still screened according to the currently set threshold; for the preferred area, the currently set confidence threshold is reduced; if suspected target information is detected in this round, calculate the position information of each suspected target, and transmit it to the ground again, then return to Step 4; If no suspected target information is detected in this round and the confidence threshold corresponding to the preferred area is greater than 0, continue to reduce the currently set confidence threshold of the preferred area and retrieve all suspected targets in the image again; if no suspected target information is detected in this round and the confidence corresponding to the preferred area is 0, then the satellite ends the tracking task.

2. The method for autonomous target tracking of a medium and high orbit remote sensing satellite based on correction of off-satellite information according to claim 1, wherein Out-of-satellite information includes but is not limited to the interpretation results obtained by manual or machine calculation of the satellite images on the ground, the imaging and calculation results from other satellites, the ground-based / space-based radar detection results, and the AIS automatic identification system signals.

3. A method for autonomous target tracking of a medium and high orbit remote sensing satellite based on correction of external information, characterized in that, In Step 5, during the continuous tracking of the target by the satellite, the trajectory and position information of the target are continuously generated, and the distance between the target and the edge of the satellite imaging field of view is calculated; If the distance between the target and the edge of the satellite imaging field of view is less than the set pointing adjustment threshold Q, then adjust the satellite imaging pointing to adjust the center of the imaging field of view to the target position, so that the target returns to the center position area of the imaging area.

4. A method for autonomous target tracking of a medium and high orbit remote sensing satellite based on correction of out-of-satellite information according to claim 3, characterized in that, The pointing adjustment threshold Q value is expressed in pixels, and the Q value is preferably 5% - 30% of the imaging field of view.

5. The method for autonomous target tracking based on off-satellite information correction of a medium and high orbit remote sensing satellite according to claim 1, wherein In Step 3, the suspected target information includes but is not limited to the target number, target type, target position and trajectory, target confidence, and target detection time.

6. The method for autonomous target tracking based on off-satellite information correction of a medium and high orbit remote sensing satellite according to claim 1, wherein In step 2, the object detection algorithms include but are not limited to Yolo, SSD, and RetinaNet; the confidence level of the object detection algorithm is a value attached to each detected result in the object detection task by the algorithm, representing the credibility of the result, and the range of the confidence level is from 0 to 1, where 0 means the algorithm does not believe the result at all, and 1 means the algorithm is certain that the result is correct.

7. A method for autonomous target tracking of a medium- and high-orbit remote sensing satellite based on correction with external information according to claim 6, characterized in that, In step 2, the initial value of the confidence level threshold of the object detection algorithm is preferably 0.3 to 0.

7.

8. A method for autonomous target tracking of a medium and high orbit remote sensing satellite based on correction of external information, characterized in that, In step 6, the preferred region setting value is preferably 2% to 20% of the imaging field of view.

9. The method for autonomous target tracking of a medium and high orbit remote sensing satellite based on correction of off-satellite information according to claim 1, characterized in that, In step 1, the task settings include imaging time, imaging interval, imaging pointing position, and task duration; among them, the imaging interval is preferably less than 10 minutes.

10. A method for autonomous target tracking of a medium and high orbit remote sensing satellite based on correction of off-satellite information according to claim 1, characterized in that, In step 5, the satellite adopts an object tracking algorithm, and by continuously processing the image imaging sequence, completes the association of the same object in multiple scene sequence images, forms the object motion trajectory, and continuously generates the trajectory and position information of the object.

Images