Onboard real-time moving target detection noise removal method and system

Abstract

The application provides a real-time moving target detection noise removing method on satellite, comprising the following steps: step S1, processing a target in real time according to frames, and inputting a preliminary detection result of each frame, which contains the position (row, col) of the target in a video image and the target number; step S2, calculating the projection point of the target on the earth surface in a fixed earth coordinate system and the projection point of the estimated height H of the target according to the position of the target in the video image; step S3, storing the projection point of the target on the earth surface and the projection point of the estimated height H of the target into a cache, and maintaining the objects including true targets and noises in the fixed earth coordinate system; and step S4, outputting the positioning information of the true target in real time according to the maintained objects. The application is simple to implement, can effectively remove the noises forming tracks on the image, and ensures the continuity of the true target.

Description

Technical Field

[0001] This invention relates to a noise removal method, and more particularly to a method for removing noise from real-time moving target detection on a satellite. Background Technology

[0002] Satellite remote sensing systems can detect and track moving targets over a large area of ​​the Earth's surface. In particular, optical imaging satellite systems are widely used in the field of satellite remote sensing and tracking due to their advantages such as low power consumption and high resolution.

[0003] By using large-area arrays and satellite attitude maneuvers, video imaging (continuous multi-frame images) of regions of interest can be achieved on low-Earth orbit satellites. For low-Earth orbit video imaging, due to the influence of satellite relative motion, Earth curvature, target distance, etc., ground targets and backgrounds may both appear to be in relative motion in the video images, while the target to be detected is the target that is moving relative to the background (Earth-fixed coordinate system).

[0004] Traditional moving target detection methods include continuous inter-frame differencing and multi-frame trajectory association. Reference 1 (Huang Zongfu, A Real-time Detection Algorithm for Weak Targets in Astronomical Optoelectronic Image Sequences, *Signal Processing*, Vol. 26, No. 9, 2010) proposes a fast target detection method using multi-frame correlation and multi-frame differencing to address the target detection and tracking needs of ground-based large-field-of-view astronomical telescopes. Reference 2 (Sima Duan, Infrared Weak Target Detection Based on Saliency and Inter-frame Difference, *Aerospace Electronic Countermeasures*, Vol. 31, No. 4, 2015) proposes a target trajectory detection method based on a combination of image saliency and inter-frame differencing for weak targets in complex backgrounds. Reference 3 (Yang Fen, Research and Hardware Design of Target Trajectory Association Detection Algorithm for Infrared Sequence Images, Graduate School of National University of Defense Technology, 2008) proposes a point target trajectory association algorithm based on linear least squares prediction for infrared sequence images, implementing target trajectory association in the sequence images. Reference 4 (Chen Xiao, A range-weighted probabilistic data association algorithm for tracking maneuvering targets, *Journal of Shanghai Jiao Tong University*, 2018, Vol. 52, No. 4) utilizes a range-weighted probabilistic data association algorithm for filtering, improving the real-time performance and accuracy of maneuvering target tracking. Reference 5 (Lin Liangkui, A mid-course ballistic target group infrared multi-sensor group tracking method based on probabilistic hypothesis density filtering, *Acta Optica Sinica*, 2011, Vol. 31, No. 2) uses motion-insensitive Kalman filtering to sequentially filter the centroid group of the target in the image plane, effectively eliminating clutter and achieving stable target tracking in the image plane.

[0005] Inter-frame differential mapping requires prior registration of consecutive frames, making it suitable for high-orbit satellites. However, for low-orbit satellites tracking targets with rapid attitude changes and large relative motion, the overlap between consecutive frames may be small, and the geometric registration model between consecutive frames may be inaccurate, leading to significant registration errors. Multi-frame trajectory correlation can remove random noise from images, but due to the relative motion during satellite imaging, some special false alarm points, such as fixed points on the ground, can also form trajectories on the image, making it difficult for random noise removal methods to filter them out.

[0006] The technical problem to be solved by the present invention is to provide a method for noise removal in real-time moving target detection on satellite, thereby reducing the false alarm rate. The key solutions are: (1) removing noise introduced by ground stationary points that will form trajectories on the image; (2) avoiding the loss of the real-time position of the target due to the temporary loss of target detection in the image, and ensuring the real-time continuous output of the true target position. Summary of the Invention

[0007] This invention adopts the following technical solution: a method for noise removal in real-time onboard moving target detection, comprising the following steps:

[0008] Step S1 processes the target in real time frame by frame and inputs the preliminary detection results for each frame, including the target's position (row, column) = (row,col) and target number in the video image;

[0009] Step S2 calculates the target's projection point on the Earth's fixed coordinate system based on the target's position in the video image. And the projection point of the estimated elevation H.

[0010] Step S3: Project the ground surface points And the projection point of the estimated elevation H. Store in the cache, and maintain objects including real targets and noise using the Earth-fixed coordinate system;

[0011] Step S4 outputs the location information of the real target in real time based on the maintained object.

[0012] Preferably, step S2 includes the following steps:

[0013] Step S2.1 Calculate the unit view vector of the target in the Earth-fixed coordinate system based on its position (row, column) = (row, col) in the video image.

[0014] Step S2.2 Based on the view vector Satellite position in Earth fixed coordinate system Earth ellipsoid model parameters, including the semi-major axis a and semi-minor axis b, are used to calculate the Earth's surface projection points in the Earth-fixed coordinate system.

[0015] Step S2.3 Let the projection point of the predicted elevation H during the iteration be... in initial value

[0016] Step S2.4 Project the points Convert the coordinates to longitude lon1, latitude lat1, and elevation h1 in a geodetic coordinate system, and calculate the projection points. The outer normal vector at the location

[0017]

[0018] Step S2.5 Calculate the projection points during the update iteration process.

[0019]

[0020] Where < > denotes the vector dot product operation.

[0021] Step S2.6 Repeat steps S2.4 to S2.5 until |h1-H| < 10- 8 m;

[0022] Step S2.7 When step S2.6 is satisfied, the projection point of the predicted elevation H is determined.

[0023] Preferably, in step S3,

[0024] Step S3.1 When maintaining objects including real targets and noise in the Earth-fixed coordinate system, determine whether there are preliminary detection results with the same target number for the maintained object. If each frame of preliminary detection results contains the same target number as the maintained object, proceed to step S3.2; if the maintained object does not have a preliminary detection result with the same target number, proceed to step S3.4.

[0025] Step S3.2 Project the corresponding Earth-fixed coordinate system surface points And the projection point of the estimated elevation H. Store the target in the cache corresponding to the target number, and after judging the target attributes by combining the target of the current frame and the information in the cache before this moment, proceed to step S3.3;

[0026] Step S3.3: Determine if there are any maintenance objects with the same target number as the detection result. If so, determine the corresponding Earth-fixed coordinate system surface projection point. And the projection point of the estimated elevation H. The new target is stored in the buffer as the first frame; if it does not exist, the location information of the real target is output directly.

[0027] Step S3.4 When there are missing preliminary detection results for objects containing both real targets and noise maintained in the Earth-fixed coordinate system, determine whether the number of recursive frames is less than the number of recursive end frames N. end If the number of recursive frames is less than the number of frames N at the end of the recursion... end At that time, the target's position is recursively calculated based on its attributes in the Earth-fixed coordinate system; if the number of recursion frames is not less than the number of recursion termination frames N. end If this happens, the object will be removed from the maintained objects list;

[0028] In step S3.5, when there is a new target, the distance of the new target is compared with the distance of the recursive result in step S3.4 and the preliminary detection result. If the distance is less than the threshold D, the target number of the new target is modified to the target number of the recursive target and maintained as an object with the same target number. Then proceed to step S3.2. If the distance is not less than the threshold D, proceed to step S3.3.

[0029] Preferably, in step S3.2, the method for determining the target attribute is as follows:

[0030] If an object has previously satisfied or currently satisfies condition 1, it is marked as noise; otherwise, it is a true target.

[0031] Condition 1 is Where angle represents the relative motion angle between the target and the satellite, and std represents all surface projection points in the cache, including the current frame. standard deviation Projection point for estimating elevation H The velocity vector estimate is given by ∪, where ∪ represents a logical OR operation and ∩ represents a logical AND operation; the relative motion angle is calculated as follows:

[0032]

[0033] in, Indicates the satellite's position in a fixed Earth coordinate system. The value represents the satellite velocity in the Earth-fixed coordinate system, and × represents the vector cross product operation.

[0034] Preferably, in step S3.4, if the current frame is the k-th frame and the preliminary detection result is missing, the position of the true target in the current frame is... speed The recursive method is

[0035]

[0036] Where k-1 represents the previous frame of the current frame, and Δt represents the time interval between the current frame and the previous frame;

[0037] When preliminary detection results are missing, the recursive method for noise estimation is as follows:

[0038] Step S3.4.1 Based on the Earth-fixed coordinate system surface projection points of the (k-1)th frame Estimate the view vector of the current frame

[0039]

[0040] Step S3.4.2 Repeat step S2 to calculate the position.

[0041] A satellite-based real-time moving target detection and noise removal system includes the following modules:

[0042] Module M1 processes the target in real time frame by frame and inputs the preliminary detection results for each frame, including the target's position (row, column) = (row,col) and target number in the video image;

[0043] Module M2 calculates the target's projection point on the Earth's fixed coordinate system based on the target's position in the video image. And the projection point of the estimated elevation H.

[0044] Module M3 will project the ground surface points And the projection point of the estimated elevation H. Store in the cache, and maintain objects including real targets and noise using the Earth-fixed coordinate system;

[0045] Module M4 outputs the location information of the real target in real time based on the maintained object.

[0046] Preferably, module M2 includes the following modules:

[0047] Module M2.1 calculates the unit view vector of the target in the Earth-fixed coordinate system based on its position (row, column) = (row, col) in the video image.

[0048] Module M2.2 based on view vectors Satellite position in Earth fixed coordinate system Earth ellipsoid model parameters, including the semi-major axis a and semi-minor axis b, are used to calculate the Earth's surface projection points in the Earth-fixed coordinate system.

[0049] Module M2.3 sets the projection point of the estimated elevation H during the iteration process as follows: in initial value

[0050] Module M2.4 will project the points Convert the coordinates to longitude lon1, latitude lat1, and elevation h1 in a geodetic coordinate system, and calculate the projection points. The outer normal vector at the location

[0051]

[0052] Module M2.5 calculates the projection points during the update iteration process.

[0053]

[0054] Where < > denotes the vector dot product operation.

[0055] Module M2.6 repeatedly triggers modules M2.4 to M2.5 to operate until |h1-H| < 10. -8 m;

[0056] When module M2.7 satisfies the conditions of module M2.6, the projection point of the predicted elevation H is determined.

[0057] Preferably, in module M3,

[0058] When maintaining objects including real targets and noise in the Earth-fixed coordinate system, module M3.1 determines whether there are preliminary detection results with the same target number for the maintained object. If the preliminary detection results of each frame contain the same target number as the maintained object, it proceeds to module M3.2; if there are no preliminary detection results with the same target number for the maintained object, it proceeds to module M3.4.

[0059] Module M3.2 projects the corresponding Earth-fixed coordinate system surface points. And the projection point of the estimated elevation H. Store the target in the cache corresponding to the target number. After judging the target attributes by combining the target of the current frame and the information in the cache before this moment, proceed to module M3.3.

[0060] Module M3.3 determines whether there are any maintenance objects with the same target number as the detection result. If so, it sets the corresponding Earth-fixed coordinate system surface projection point. And the projection point of the estimated elevation H. The new target is stored in the buffer as the first frame; if it does not exist, the location information of the real target is output directly.

[0061] When there are missing preliminary detection results for objects containing both real targets and noise maintained in the Earth-fixed coordinate system, module M3.4 determines whether the number of recursive frames is less than the number of recursive end frames N. end If the number of recursive frames is less than the number of frames N at the end of the recursion... end At that time, the target's position is recursively calculated based on its attributes in the Earth-fixed coordinate system; if the number of recursion frames is not less than the number of recursion termination frames N. end If this happens, the object will be removed from the maintained objects list;

[0062] When module M3.5 contains a new target, it compares the distance between the new target and the preliminary detection result based on the recursive result of module M3.4. If the distance is less than the threshold D, the target number of the new target is modified to the target number of the recursive target, and the same maintenance is performed as for objects with the same target number, and then it enters module M3.2; if the distance is not less than the threshold D, it enters module M3.3.

[0063] Preferably, in module M3.2, the target attribute determination method is as follows:

[0064] If an object has previously satisfied or currently satisfies condition 1, it is marked as noise; otherwise, it is a true target.

[0065] Condition 1 is Where angle represents the relative motion angle between the target and the satellite, and std represents all surface projection points in the cache, including the current frame. standard deviation Projection point for estimating elevation H The velocity vector estimate is given by ∪, where ∪ represents a logical OR operation and ∩ represents a logical AND operation; the relative motion angle is calculated as follows:

[0066]

[0067] in, Indicates the satellite's position in a fixed Earth coordinate system. The value represents the satellite velocity in the Earth-fixed coordinate system, and × represents the vector cross product operation.

[0068] Preferably, in module M3.4, if the current frame is the k-th frame and the preliminary detection results are missing, the position of the true target in the current frame is... speed The recursive method is

[0069]

[0070] Where k-1 represents the previous frame of the current frame, and Δt represents the time interval between the current frame and the previous frame;

[0071] When preliminary detection results are missing, the recursive method for noise estimation is as follows:

[0072] Module M3.4.1 projects the Earth's surface points in the Earth-fixed coordinate system based on the (k-1)th frame. Estimate the view vector of the current frame

[0073]

[0074] Module M3.4.2 repeatedly triggers module M2 to work and calculates the position.

[0075] The positive and progressive effects of this invention are as follows:

[0076] 1. This invention provides a method for noise removal in real-time on-board moving target detection, which transforms the target motion trajectory in the image plane coordinate system of a video image to the Earth-fixed coordinate system for calculating motion attributes.

[0077] 2. This invention utilizes the motion law of targets in the Earth-fixed coordinate system to eliminate stationary targets in the Earth-fixed coordinate system on the one hand; on the other hand, by recursively extrapolating the position in the Earth-fixed coordinate system, it achieves the maintenance of the position continuity of real targets, ensuring that the target can output position information in real time for each cycle.

[0078] 3. The method of the present invention is reasonable, simple to calculate and easy to implement. It can effectively remove noise that forms a trajectory on the image and ensure the continuity of the true target. Attached Figure Description

[0079] Figure 1 This is a schematic diagram of the process of the present invention;

[0080] Figure 2 This is a schematic diagram of the preliminary detection results of the image plane coordinate system over a period of time.

[0081] Figure 3 This is a schematic diagram showing the result of noise removal using the method of the present invention. Detailed Implementation

[0082] The preferred embodiments of the present invention are given below with reference to the accompanying drawings to illustrate the technical solution of the present invention in detail.

[0083] Considering that during the process of low-orbit satellites achieving video imaging of the region of interest, the target motion model in the image plane coordinate system is affected by the target motion model in the Earth-fixed coordinate system, the satellite's position and velocity, the satellite's attitude, and the camera's imaging geometry, the noise removal method of this invention adopts a method of analyzing the target motion attributes in the Earth-fixed coordinate system.

[0084] The ultimate goal of satellite-based target detection is to determine the target's position in a fixed Earth coordinate system. Single-satellite detection requires prior knowledge of the target's altitude for localization. The target's (including noise) position in the video image (row, column) = (row, col) can be used to calculate its projection point on the Earth's surface in a fixed Earth coordinate system. And the projection point of the estimated elevation H.

[0085] First, calculate the unit view vector of the target (including noise) in the Earth-fixed coordinate system based on its position (row, column) = (row, col) in the video image.

[0086] Based on view vector Satellite position in Earth fixed coordinate system Earth ellipsoid model parameters (semi-major axis a, semi-minor axis b), calculate the Earth surface projection points in the Earth-fixed coordinate system.

[0087] Let the projection point of the predicted elevation H during the iteration process be... initial value

[0088] Projection point Convert the coordinates to longitude lon1, latitude lat1, and elevation h1 in a geodetic coordinate system, and calculate the projection points. The outer normal vector at the location

[0089]

[0090] Projection points during the update and iteration process

[0091]

[0092] Where < > denotes the vector dot product operation.

[0093] Repeat the iterations corresponding to Formula 2 until |h1-H|<10 -8 m (reaching the convergence threshold).

[0094] The final estimated elevation H projection point

[0095] By projecting noisy target points onto a fixed Earth coordinate system, their motion characteristics over a period of time can be analyzed. For noise at a fixed point in the fixed Earth coordinate system, the direction of its angular momentum is close to that of the satellite's motion; therefore, the relative motion angle can be used as one of the criteria for target attributes. Simultaneously, the standard deviation of its projection points onto the Earth ellipsoid model can be used as another criterion for target attributes. Furthermore, the velocity of a specific target in the fixed Earth coordinate system generally has a threshold (e.g., a maximum value v). max ).

[0096] In summary, if an object has previously satisfied or currently satisfies condition 1, it is marked as noise; otherwise, it is a true target. Condition 1 is... Where angle represents the relative motion angle between the target and the satellite, and std represents all surface projection points in the cache, including the current frame. standard deviation Projection point for estimating elevation H The velocity vector estimate is given by ∪, where ∪ represents a logical OR operation and ∩ represents a logical AND operation. The relative motion angle is calculated using the following method:

[0097]

[0098] in, Indicates the satellite's position in a fixed Earth coordinate system. The value represents the satellite velocity in the Earth-fixed coordinate system, and × represents the vector cross product operation.

[0099] If the current frame is the k-th frame and the initial detection results are missing, the location of the true target in the current frame... speed The recursive method is

[0100]

[0101] Where k-1 represents the previous frame of the current frame, and Δt represents the time interval between the current frame and the previous frame.

[0102] When preliminary detection results are missing, the noise estimation method is as follows: based on the Earth-fixed coordinate system surface projection points of the (k-1)th frame. Estimate the view vector of the current frame

[0103]

[0104] Current position Similarly, the calculation can be performed.

[0105] Considering that both real targets and noise may go undetected for a short period, it's necessary to simultaneously compare the distance between the recursive value and the newly added preliminary detection results during object maintenance. Potentially changed targets should be modified to have the same target number. When objects maintained in the Earth-fixed coordinate system (including real targets and noise) have missing preliminary detection results, their positions are recursively calculated based on their attributes in the Earth-fixed coordinate system. The distance between the recursive result and any newly added targets (if any) in the preliminary detection results is compared. If the distance is less than a threshold D, the target number of the newly added target is modified to the recursive target's target number, and it is treated as an object with the same target number and maintained in the same way.

[0106] The effectiveness of the method of the present invention will be verified below using target detection data from a certain satellite. Figure 2 The image shows targets that can form trajectories detected in the image plane coordinate system over a period of time. In this coordinate system, it is impossible to effectively distinguish between real targets and noise. Both real targets and noise appear as speckles / arcs in the image plane coordinate system. Using the method of this invention, these targets are processed in real time, and the results after differentiation in the Earth-fixed coordinate system are as follows. Figure 3 As shown, noise has been effectively removed, and the position of the true target remains continuous over a period of time.

[0107] The present invention also provides an on-board real-time moving target detection noise removal system, which can be implemented by executing the process steps of the on-board real-time moving target detection noise removal method. That is, those skilled in the art can understand the on-board real-time moving target detection noise removal method as a preferred embodiment of the on-board real-time moving target detection noise removal system.

[0108] A real-time moving target detection and noise removal system for satellites includes the following modules: Module M1 processes targets frame by frame in real time and inputs the preliminary detection results for each frame, including the target's position (row, column) = (row,col) and target number in the video image; Module M2 calculates the target's projection point on the Earth's fixed coordinate system based on the target's position in the video image. And the projection point of the estimated elevation H. Module M3 will project the ground surface points And the projection point of the estimated elevation H. The data is stored in the cache, and the Earth-fixed coordinate system is used to maintain objects including real targets and noise. Module M4 outputs the positioning information of real targets in real time based on the maintained objects.

[0109] Specifically, module M2 includes the following modules: Module M2.1 calculates the unit view vector of the target in the Earth-fixed coordinate system based on the target's position (row, column) = (row,col) in the video image. Module M2.2 based on view vectors Satellite position in Earth fixed coordinate system Earth ellipsoid model parameters, including the semi-major axis a and semi-minor axis b, are used to calculate the Earth's surface projection points in the Earth-fixed coordinate system. Module M2.3 sets the projection point of the estimated elevation H during the iteration process as follows: in initial value Module M2.4 will project the points Convert the coordinates to longitude lon1, latitude lat1, and elevation h1 in a geodetic coordinate system, and calculate the projection points. The outer normal vector at the location

[0110]

[0111] Module M2.5 calculates the projection points during the update iteration process.

[0112]

[0113] Where < > represents the vector inner product operation, module M2.6 repeatedly triggers modules M2.4 to M2.5 to work until |h1-H|<10. -8 m; Module M2.7 When module M2.6 is satisfied, the projection point of the predicted elevation H.

[0114] Specifically, in module M3, when module M3.1 maintains objects including real targets and noise in the Earth-fixed coordinate system, it determines whether there are preliminary detection results with the same target number for the maintained object. If each frame's preliminary detection results contain a target number identical to the maintained object, it proceeds to module M3.2; if the maintained object does not have a preliminary detection result with the same target number, it proceeds to module M3.4. Module M3.2 projects the corresponding Earth-fixed coordinate system surface points... And the projection point of the estimated elevation H. The data is stored in the cache corresponding to the target number. After judging the target attributes by combining the target in the current frame and the information in the cache before this moment, it enters module M3.3. Module M3.3 checks whether there is a maintained object with the same target number that the detection result does not have. If it does, it projects the corresponding Earth-fixed coordinate system surface point. And the projection point of the estimated elevation H. The new target is stored in the buffer as the first frame; if it does not exist, the location information of the real target is directly output; when there are missing preliminary detection results for objects containing real targets and noise maintained in the Earth-fixed coordinate system, module M3.4 determines whether the number of recursive frames is less than the number of recursive end frames N. end If the number of recursive frames is less than the number of frames N at the end of the recursion...end At that time, the target's position is recursively calculated based on its attributes in the Earth-fixed coordinate system; if the number of recursion frames is not less than the number of recursion termination frames N. end If the target is not found in the target list, the target will be removed from the maintenance list. When module M3.5 contains a new target, it will compare the distance between the new target and the target in the preliminary detection result based on the recursive result of module M3.4. If the distance is less than the threshold D, the target number of the new target will be changed to the target number of the recursive target and treated as an object with the same target number. Then it will be maintained in the same way and enter module M3.2. If the distance is not less than the threshold D, it will enter module M3.3.

[0115] Specifically, in module M3.2, the target attribute judgment method is as follows: when an object has previously satisfied or the current frame satisfies condition 1, it is marked as noise; otherwise, it is a true target; condition 1 is... Where angle represents the relative motion angle between the target and the satellite, and std represents all surface projection points in the cache, including the current frame. standard deviation Projection point for estimating elevation H The velocity vector estimate is given by ∪, where ∪ represents a logical OR operation and ∩ represents a logical AND operation; the relative motion angle is calculated as follows:

[0116]

[0117] in, Indicates the satellite's position in a fixed Earth coordinate system. The value represents the satellite velocity in the Earth-fixed coordinate system, and × represents the vector cross product operation.

[0118] Specifically, in module M3.4, if the current frame is the k-th frame and the initial detection results are missing, the position of the true target in the current frame is determined. speed The recursive method is

[0119]

[0120] Where k-1 represents the previous frame of the current frame, and Δt represents the time interval between the current frame and the previous frame; when the preliminary detection results are missing, the noise recursion method is as follows: Module M3.4.1 calculates the noise level based on the Earth-fixed coordinate system surface projection points of the (k-1)th frame. Estimate the view vector of the current frame

[0121]

[0122] Module M3.4.2 repeatedly triggers module M2 to work and calculates the position.

[0123] Those skilled in the art will understand that, besides implementing the system and its various devices, modules, and units provided by this invention in the form of purely computer-readable program code, the same functions can be achieved entirely through logical programming of the method steps, enabling the system and its various devices, modules, and units to function in the form of logic gates, switches, application-specific integrated circuits, programmable logic controllers, and embedded microcontrollers. Therefore, the system and its various devices, modules, and units provided by this invention can be considered a hardware component, and the devices, modules, and units included therein for implementing various functions can also be considered structures within the hardware component; alternatively, the devices, modules, and units for implementing various functions can be considered both software modules implementing the method and structures within the hardware component.

[0124] Specific embodiments of the present invention have been described above. It should be understood that the present invention is not limited to the specific embodiments described above, and those skilled in the art can make various changes or modifications within the scope of the claims, which do not affect the essence of the present invention. Unless otherwise specified, the embodiments and features described in this application can be arbitrarily combined with each other.

Claims

1. A method for noise removal in real-time moving target detection on board a satellite, characterized in that, Includes the following steps: Step S1 processes the target in real time frame by frame and inputs the preliminary detection results for each frame, including the target's position (row, column) = (row,col) and target number in the video image; Step S2 calculates the projection point of the target on the earth-fixed coordinate system surface according to the position of the target in the video image and the projection point of the estimated height H Step S3 projects the surface point and the projected point of the estimated height H into the buffer, which is maintained in the Earth-fixed coordinate system, containing the true target and noise objects Step S4 outputs the location information of the real target in real time based on the maintained object; In step S3, Step S3.1 When maintaining objects including real targets and noise in the Earth-fixed coordinate system, determine whether there are preliminary detection results with the same target number for the maintained object. If each frame of preliminary detection results contains the same target number as the maintained object, proceed to step S3.2; if the maintained object does not have a preliminary detection result with the same target number, proceed to step S3.

4. Step S3.2 projects the corresponding terrestrial fixed coordinate system surface projection point and the projection point of the estimated height H into the cache corresponding to the target number, and after target attribute judgment combining the current frame target and the information in the cache before this time, enters step S3.3; Step S3.3 judges whether there is a maintenance object with the same target number and the same detection result, and if so, the corresponding earth-fixed coordinate system surface projection point and the projection point of the estimated height H is stored in the buffer as the first frame of a new target If it does not exist, the location information of the real target will be output directly; Step S3.4, when the initial detection result of the object containing true target and noise maintained in the earth-fixed coordinate system is missing, judging whether the recursive frame number is less than the recursive end frame number N end , if the recursive frame number is less than the recursive end frame number N end , recursively calculating the position of the object in the earth-fixed coordinate system according to the target attribute; if the recursive frame number is not less than the recursive end frame number N end , removing the object from the maintained objects. In step S3.5, when there is a new target, the distance of the new target is compared with the distance of the recursive result in step S3.4 and the preliminary detection result. If the distance is less than the threshold D, the target number of the new target is modified to the target number of the recursive target and maintained as an object with the same target number. Then proceed to step S3.

2. If the distance is not less than the threshold D, proceed to step S3.

3. In step S3.2, the method for determining the target attribute is as follows: If an object has previously satisfied or currently satisfies condition 1, it is marked as noise; otherwise, it is a true target. The condition 1 is where angle represents the relative motion angle between the target and the satellite, std represents the standard deviation of all ground projection points in the buffer including the current frame, The velocity vector estimate of the projection point with the estimated height H height, The velocity vector estimate of the projection point with the estimated height H height, The logical OR operation, and the logical AND operation; The relative motion angle angle calculation method is: wherein denotes the satellite position in the earth fixed coordinate system, denotes the satellite velocity in the earth fixed coordinate system, x denotes the vector cross product operation.

2. The on-board real-time moving target detection noise removal method according to claim 1, characterized in that, Step S2 includes the following steps: Step S2.1 calculates the unit view vector of the target in the Earth fixed coordinate system according to the position (row, col) of the target in the video image Step S2.2 determining the view vector Satellite position in earth fixed coordinate system Earth ellipsoid model parameters, wherein the earth ellipsoid model parameters comprise a semi-major axis a, a semi-minor axis b, calculating a surface projection point of the earth fixed coordinate system Step S2.3 lets the projected point of the estimated height H in the iteration process be wherein The initial value is Step S2.4 converts the projection point into geodetic coordinates lonl, latl, h1, calculates the outward normal vector at the projection point Step S2.5 calculates the projection point in the update iteration process Where <> represents the vector dot product operation, Step S2.6 repeats Step S2.4 - Step S2.5 until |h1-H| < 10 -8 m; Step S2.7 When step S2.6 is satisfied, the projection point of the estimated height H height 3. The on-board real-time moving target detection noise removal method according to claim 1, characterized in that, In step S3.4, if the current frame is the kth frame and the preliminary detection result is missing, the position of the true target in the current frame is determined according to the following recursive method: velocity The recursive method is Where k-1 represents the previous frame of the current frame, and Δt represents the time interval between the current frame and the previous frame; When preliminary detection results are missing, the recursive method for noise estimation is as follows: Step S3.4.1 Earth-fixed coordinate system surface projection point according to the k-1 frame Estimate the view vector of the current frame Step S3.4.2 repeats step S2 to calculate the position 4. An on-board real-time moving target detection noise removal system, characterized in that, Includes the following modules: Module M1 processes the target in real time frame by frame and inputs the preliminary detection results for each frame, including the target's position (row, column) = (row,col) and target number in the video image; The module M2 calculates the projection point of the target on the earth fixed frame surface projection point and the estimated height H height projection point Module M3 stores the ground projection point and the projection point of the estimated height H in the cache, maintaining the objects in the Earth-fixed coordinate system, including true targets and noise. Module M4 outputs the location information of the real target in real time based on the maintained object; In module M3, When maintaining objects including real targets and noise in the Earth-fixed coordinate system, module M3.1 determines whether there are preliminary detection results with the same target number for the maintained object. If the preliminary detection results of each frame contain the same target number as the maintained object, it proceeds to module M3.2; if there are no preliminary detection results with the same target number for the maintained object, it proceeds to module M3.

4. Module M3.2 projects the corresponding terrestrial fixed coordinate system ground projection point and the projection point of the estimated height H is stored in the cache corresponding to the target number, and after target attribute judgment combining the current frame target and the information in the cache before this time, it enters module M3.3; Module M3.3 judges whether there is a maintenance object with the same target number of detection results, and if so, the projection point of the earth-fixed coordinate system on the ground corresponding to the maintenance object and the projection point of the estimated height H is stored in the buffer as a new target If it does not exist, the location information of the real target will be output directly; Module M3.4 judges whether the recursive frame number is less than the recursive end frame number N when the preliminary detection result of the object containing true target and noise maintained in the earth-fixed coordinate system is missing end If the recursive frame number is less than the recursive end frame number N end , the position of the object is recursively calculated according to the target attribute in the earth-fixed coordinate system; if the recursive frame number is not less than the recursive end frame number N end , the object is removed from the maintained objects. When module M3.5 contains a new target, it compares the distance between the new target and the preliminary detection result based on the recursive result of module M3.

4. If the distance is less than the threshold D, the target number of the new target is modified to the target number of the recursive target, and the same maintenance is performed as for objects with the same target number, and then it proceeds to module M3.2; if the distance is not less than the threshold D, it proceeds to module M3.

3. In module M3.2, the method for determining the target attribute is as follows: If an object has previously satisfied or currently satisfies condition 1, it is marked as noise; otherwise, it is a true target. The condition 1 is where angle represents the relative motion angle between the target and the satellite, std represents the standard deviation of all ground projection points in the buffer including the current frame, The velocity vector estimate of the projection point with the estimated height H, The velocity vector estimate of the projection point with the estimated height H, The relative motion angle angle is calculated as follows: wherein denotes the satellite position in the earth fixed coordinate system, denotes the satellite velocity in the earth fixed coordinate system, x denotes the vector cross product operation.

5. The on-board real-time moving target detection noise removal system of claim 4, wherein, Module M2 includes the following modules: Module M2.1 computes its unit view vector in the Earth fixed coordinate system as a function of its position in the video image (row, col) Module M2.2 according to the view vector Satellite position in earth fixed coordinate system Earth ellipsoid model parameters, wherein the earth ellipsoid model parameters comprise a semi-major axis a, a semi-minor axis b, a calculation of a surface projection point of the earth fixed coordinate system The module M2.3 orders the projection of the point of the iteration process with the estimated height H to the height wherein The initial value is The module M2.4 converts the projection point into geodetic coordinates lon1, lat1, h1 and calculates the outer normal vector at the projection point The module M2.5 calculates the projection point in the update iteration process Where < > denotes the vector dot product operation. Module M2.6 repeats the work of module M2.4 - module M2.5 until |hl - H| < 10 -8 m; Module M2.7 Projects the point of the estimated height H when module M2.6 is satisfied 6. The on-board real-time moving target detection noise removal system of claim 4, wherein, In the module M3.4, if the current frame is the kth frame and the preliminary detection result is missing, the position of the true target in the current frame is velocity The recursive method is Where k-1 represents the previous frame of the current frame, and Δt represents the time interval between the current frame and the previous frame; When preliminary detection results are missing, the recursive method for noise estimation is as follows: Module M3.4.1 estimates the earth-fixed coordinate system surface projection point according to the k-1th frame estimates the view vector of the current frame Module M3.4.2 repeats the triggering of module M2 to calculate the position

Images