CN110617802A - Satellite-borne moving target detection and speed estimation method - Google Patents

Abstract

A method for detecting and estimating the speed of a satellite-borne moving target, comprising the steps of: first, for each frame of a multi-frame image continuously shot: (1) processing the original star map to obtain a background-free image; (2) using an automatic The adaptive threshold method performs threshold segmentation on the background image, and stores all pixel coordinates and gray levels higher than the threshold; (3) extracts potential targets by four-connected domain targets on the saved pixels; (4) uses the area and length of the connected domain to extract potential targets. The width ratio threshold is used to eliminate false targets and calculate the centroid of each remaining connected domain; (5) Use the coordinate system transformation matrix provided by the onboard computer and the camera attitude information to project the centroid of each connected domain into the GPS coordinate system. For the connected domain extracted from each frame, the false target is eliminated, and the motion speed and direction of the real moving target in the GPS coordinate system are calculated. The method can effectively eliminate the background interference, realize the real-time detection of the moving target on the satellite and provide the geographical coordinates and movement speed of the target.

Description

A Spaceborne Moving Target Detection and Velocity Estimation Method

technical field

The present invention relates to an image processing method for multi-frame detection of moving targets, in particular to a method for real-time moving target detection and velocity estimation of a spaceborne early warning camera in orbit, which is the technical field of moving target image processing.

Background technique

On-orbit real-time detection and tracking of moving objects of interest in the atmosphere is one of the important tasks of earth observation and remote sensing satellites. Extract the target, calculate the corresponding position of the target and the size and direction of the corresponding movement speed, and then use the calculation results to track and shoot the target for a long time, complete the analysis of the target and report the target information to the ground in time to achieve rapid detection and tracking.

Due to the long imaging distance, the area of the target on the image ranges from a few to more than a dozen pixels, there is no internal texture feature, and there is less external shape information. , the false alarm rate of the algorithm is generally high. The algorithm usually preprocesses a single frame of images to improve the signal-to-noise ratio of suspicious targets, then performs threshold segmentation on the images and extracts connected domains; and then uses the processing results of multiple frames of images to detect moving targets.

The design of traditional moving target processing algorithms is under ideal conditions, ignoring satellite motion and camera attitude changes, and directly performing multi-frame matching of phase coordinates on the preprocessed image. This method cannot be applied to the real spaceborne environment. Down. At the same time, most of the methods such as bilateral filtering, wavelet transform and other algorithms require a large amount of calculation and cannot be calculated in real time on-orbit.

Due to the direct use of phase coordinates for multi-frame target detection, the traditional methods are inaccurate for the setting of velocity magnitude and velocity direction threshold: most methods directly use the same connected domain between multiple frames to displace on the phase surface to complete the velocity magnitude estimation. Using the displacement direction vector to complete the estimation of the velocity direction, this method simply uses the theoretical pixel resolution of the camera, but ignores the optical axis direction of the camera, the influence of the satellite's attitude on the camera's shooting angle to the ground, and the pixel resolution. Therefore, the detection false alarm rate of traditional methods is generally higher.

SUMMARY OF THE INVENTION

The technical problem solved by the present invention is: to overcome the deficiencies of the prior art, to provide a spaceborne moving target detection and speed estimation method, and to solve the problem that the spaceborne camera cannot detect the ground moving target in real time and provide the size and speed of the target moving speed.

The technical scheme solved by the present invention is: a method for detecting and estimating the speed of a satellite-borne moving target, and the steps are as follows:

(1) Background suppression processing is performed on the continuously shot multi-frame satellite-borne Earth observation images to obtain background-removed images;

(2) Perform threshold segmentation on the background-removed image in step (1), and then perform connected domain extraction to remove some false target connected domains to obtain effective connected domains;

Specifically, the adaptive threshold method is used to perform threshold segmentation on the background image, and the coordinates and gray values of all pixels higher than the threshold are stored; The area and aspect ratio thresholds of the domain are used to eliminate some false target connected domains to obtain effective connected domains.

(3) Extract the centroid of the effective connected domain in step (2), and obtain the precise coordinates of the centroid of the effective connected domain on the phase plane of the photographing camera.

(4) Utilize the coordinate transformation matrix and the camera optical axis pointing parameter to complete the coordinate transformation from the precise coordinates of the centroid on the phase plane of the shooting camera to the GPS coordinate system, and obtain the effective connected domain centroid coordinates under the GPS coordinate system;

(5) According to the centroid coordinates of the effective connected domain in the GPS coordinate system, the geographic location corresponding to the centroid of the corresponding effective connected domain is obtained, and the corresponding movement speed and direction are calculated, and the method of extracting the real target by comparing the connected domain of multiple frames of images is used. Extract the connected domain corresponding to the real moving target, finally realize the moving target detection, and realize its speed estimation.

Preferably, the multiple frames of satellite-borne earth observation images continuously captured in step (1) are: an early warning camera installed on an orbiting satellite with an orbital height of 500 km, and its field of view angle is 4°.

Preferably, in step (1), background suppression processing is performed on the consecutively shot multi-frame satellite-borne Earth observation images to obtain a background-removed image. Specifically, the template-based morphological Top-Hat algorithm is performed on each frame of image to obtain Go to background image.

Preferably, the extraction of connected domains includes: roughly extracting all potential target connected domains based on the four-connected domain method, calculating the area of each connected domain and the maximum length-width ratio of the connected domains and comparing them with a set fixed threshold, and eliminating those that do not meet the requirements. The false target connected domain is obtained, and the remaining connected domain is an effective connected domain.

Preferably, the phase coordinates of the center of mass of the Unicom domain are converted into the coordinates of the GPS coordinate system, specifically: using the transformation matrix C from the camera body coordinate system to the GPS coordinate system, and the displacement of the satellite and the earth's center at this moment Representation in the GPS coordinate system, corresponding to the radius of the earth R; the direction vector of the optical path corresponding to the precise coordinates of the connected domain in the phase plane in the camera body coordinate system Calculated by the following formula:

in the formula is the direction vector of the light path corresponding to the pixel in the GPS coordinate system; it is calculated by the following formula:

in the formula is the direction vector of the object corresponding to the connected domain relative to the center of the earth in the GPS coordinate system, h is the height of the target, and r is the distance from the satellite to the object corresponding to the connected domain. According to the above formula, the representation of r and direction vector in the GPS coordinate system can be solved

Preferably, the method for extracting the real target by comparing the connected domains of multiple frames of images is specifically:

(1) The connected domains extracted from each frame of image are sorted according to the area and energy of the connected domains. Specifically, first sort according to the area of the connected domain, and place the one with the largest area at the front of the queue. For the connected domain with the same area, compare the sum of the gray values of its pixels, and sort by size;

(2) Starting from the second frame, the connected domains are matched with the connected domains in all previous frames one by one frame by frame, and the successfully matched connected domain combinations are stored in the corresponding connected domains of the current frame;

Preferably, the connected domains are matched, specifically: extracting the connected domains one by one in the current frame and performing connectivity detection with the matched connected domains (that is, all the result connected domains before the current frame) one by one, specifically, including:

(1) When the length of the connected domain to be matched is 1, use the displacement of the two in the GPS coordinate system and the difference between the frame where the current connected domain is located and the frame where the connected domain to be matched is located to jointly calculate the motion speed of the corresponding target, and compare it with The speed threshold is compared. If it is not within the threshold range, the matching fails, and the next connected domain detection is performed; if the speed is within the threshold range, the two connected domains are merged and stored in the current connected domain. The length of the connected domain is increased by 1 , store the size of the corresponding motion speed, use the displacement of the two in the GPS coordinate system, calculate the motion direction vector and store it, and delete the data of the matched connected domain at the original position.

(2) When the length of the connected domain to be matched is greater than 1, use the method similar to (1) to find the corresponding motion velocity and the motion direction vector of the connected domain of the current connected domain and the last frame of the connected domain to be matched, and detect whether the velocity is large or not. If the threshold condition is met, use the velocity direction threshold detection to determine whether the velocity direction satisfies the threshold condition, if not at the same time, perform the next connected domain detection; The weighted calculation is stored in the current connected domain information, and the two connected domains are merged and stored in the current connected domain. data.

Preferably, the speed direction threshold value detection is: using the current frame and the speed direction vector of the matched connected domain and the corresponding speed direction vector stored in the matched connected domain to perform a dot product operation, if the calculation result is greater than the threshold value, then the threshold value detection is passed; Otherwise it fails.

The advantages of the present invention compared with the prior art are:

(1) The present invention has carried out a targeted design on the template based on the size of the moving target on the phase plane, and the algorithm precision is high;

(2) The present invention converts the phase coordinates of the centroid of the connected domain into the GPS coordinate system through the conversion matrix of the camera body coordinate system to the GPS coordinate system and the vector of the satellite relative to the earth provided by the on-board computer, and can calculate more accurately The speed and direction of the target movement;

(3) The present invention uses two thresholds of velocity magnitude and velocity direction to perform connected domain matching, and switches the matching method according to the length of the connected domain, which can greatly reduce the false alarm rate;

(4) The calculation amount of all the algorithms used in the present invention is small, and the real-time processing can be completed under the hardware conditions of the spaceborne camera, which has high practical value;

(5) The present invention accurately extracts the suspicious target area under the condition of small calculation amount, and can provide more accurate calculation results of speed and direction, thereby reducing the false alarm rate and improving the detection success rate.

Description of drawings

Figure 1 is a flow chart of a method for on-board moving target detection and velocity estimation;

Figure 2 is a specific template of the Top-Hat algorithm;

Fig. 3 is a schematic diagram of solving the direction vector of the corresponding connected domain relative to the earth's center;

FIG. 4 is a schematic diagram of the principle of connectivity judgment of a connected domain.

Detailed ways

The present invention will be further described in detail below with reference to the accompanying drawings.

A method for detecting and estimating the speed of a satellite-borne moving target according to the present invention includes the following steps: first, for each frame of a multi-frame image continuously shot: (1) for the original star map (that is, the original image, the multiple frames of continuous shooting) (2) Use adaptive threshold method to perform threshold segmentation on the background-removed image, and store all pixel coordinates and gray levels higher than the threshold; Pixels are used for four-connected domain targets to extract potential targets; (4) Use the connected domain area and aspect ratio threshold to eliminate false targets and calculate the centroid of each remaining connected domain; (5) Use the coordinate system transformation matrix provided by the onboard computer, and the camera Attitude information projects the centroids of each connected domain into the GPS coordinate system. Secondly, for the connected domain extracted from each frame, based on the motion speed and motion direction, the false target is eliminated by multi-frame comparison, and the motion speed and motion direction of the real moving target in the GPS coordinate system are calculated. The method can effectively eliminate the background interference, realize the real-time detection of the moving target on the satellite and provide the geographical coordinates and movement speed of the target.

The invention can be used in a spaceborne early warning camera to realize real-time on-orbit moving target detection and speed estimation, and provide suspicious target information to satellites in real time. According to the present invention, the amount of image data captured by the existing payload camera for reconnaissance remote sensing satellites is large, and real-time on-orbit processing cannot be realized. Therefore, the suspicious moving targets captured by the present invention cannot be provided in real time, and data transmission to the ground for processing is usually impossible. Get the current action and purpose of the moving target. The present invention can realize on-orbit detection of moving targets, provide time-sensitive information, and enhance the early warning capability of reconnaissance satellites. The concrete operation steps of the present invention are as follows:

(1) Background suppression processing is performed on the continuously shot multi-frame satellite-borne earth observation images (that is, the original images) to obtain a background-removed image, which is: in the process of continuous shooting of the ground by the camera, the current frame image is preferably subjected to background suppression and background suppression. Noise and target contrast are improved, and the background suppression processing is performed on the continuous shooting of multiple frames of satellite-borne earth observation images;

As shown in Figure 2, it is preferable to use Perform background suppression processing to obtain a background-removed image, where TH(f) is the result image (ie, background-removed image) matrix, and f is the original image matrix.

details as follows:

The first step: find the maximum image: perform the morphological expansion operation based on the ring structure element A on the original image, is the expansion operation symbol, specifically, the expansion operation includes: replacing the gray value of the corresponding pixel with the maximum gray value of the pixel in the annular area corresponding to the structural element A with the pixel as the center.

The preferred solution determined by the structural element A is to set the internal structural element Bi as a 3*3 flat structural element (that is, a 3*3 matrix, each element in the matrix is 1), and the external structural element B ₀ is a 5*5 flat structure element (that is, a 5*5 matrix, each element in the matrix is 1), take the center of the two as the origin, perform pseudo-subtraction of the matrix, and obtain the result A=B ₀ -B _i , the result A is a ring structure element, then The structure of A is shown in Figure 2, the size is 5*5, the inner 3*3 area is empty, and it is in the shape of a ring.

The second step: find the smallest image: perform the morphological corrosion operation based on the flat structure element B _i on the largest image obtained in the first step, where Θ is the corresponding corrosion operation. Specifically, the corrosion operation includes: converting the gray value of the corresponding pixel Replaced with the pixel as the center, the structure element B _i corresponds to the minimum gray value of the pixel in the 3*3 area.

The third step: find the result image (ie, the background image is removed): make the difference between the original image and the minimum image obtained in the second step, that is, perform the gray value subtraction operation on each corresponding pixel of the two.

The fourth step: find the background image: set the gray value of the pixel whose gray value is less than 0 in the result image obtained in the third step to 0, that is, the background image is obtained;

(2) Perform threshold segmentation on the background image, and then extract the connected domain. details as follows:

The background-removed image is segmented by the adaptive threshold method, and the phase coordinates and the gray value of the pixels whose gray value is higher than the threshold value in the image are saved. The adaptive threshold method includes: calculating the global threshold, and its calculation formula is V _th =μ+α·σ, where μ is the gray mean value of the background-removed image, σ is the global gray standard deviation, and α is a set fixed coefficient , determines the sensitivity of the extraction target (ie detection rate), and takes into account the sensitivity and false alarm rate requirements, preferably 2 to 5, preferably α=3, to improve the balance detection rate and false alarm rate.

The four-connected domain is extracted by the clustering method for the pixels that pass the threshold method, and it is preferable to save all the connected domains with an area greater than 2 and less than 10 (due to the long shooting distance, it is considered that the moving target is no different from the point target, due to the effect of micro defocusing). , the target size is the set threshold, the user can set it according to the parameters of the camera optical system), that is, the effective connected domain.

(3) Extract the centroid of the effective connected domain, and obtain its precise coordinates on the photographing camera phase. details as follows:

In the present invention, the camera is preferably installed on a satellite with an orbital height of 500km from the ground, and is connected to the satellite through a pan/tilt, which can rotate in three axes relative to the satellite. Therefore, there is a transformation matrix C _cs from the camera body coordinate system to the satellite body coordinate system, The camera's field of view is θ, and the preferred value is 4°. The coordinates of the centroid on the phase surface are (x _i , y _i ), that is, the exact coordinates on the phase surface of the camera, where x _i is the image column direction coordinate, and y _i is Image row direction coordinates.

(4) Using the transformation matrices of multiple coordinate systems and the direction vector of the optical axis of the camera to complete the coordinate transformation of the coordinates of the centroid on the camera phase plane to the GPS coordinate system, and obtain the effective connected domain centroid coordinates in the GPS coordinate system. details as follows:

The transformation matrix C from the camera body coordinate system to the GPS coordinate system and the coordinates (x _i , y _i ) of the centroid of each connected domain on the phase surface obtained in (3), and the coordinates of each connected domain in the GPS coordinate system are obtained. The transformation matrix C from the camera body coordinate system to the GPS coordinate system consists of three transformation matrices: the transformation matrix C _Ie from the geocentric inertial coordinate system to the GPS coordinate system, and the transformation matrix from the satellite body coordinate system to the geocentric inertial coordinate system. C _sI , from the camera body coordinate system to the satellite body coordinate system C _cs . The optical axis of the camera points to the positive z-axis of the camera body coordinate system, the row direction of the image plane is parallel to the x-axis of the camera body coordinate system, the column direction of the image plane is parallel to the y-axis of the camera body coordinate system, and the camera field of view is θ. Resolution u*v, further, the preferred coordinate transformation method is as follows:

The first step: using the multiple transformation matrices obtained by the satellite attitude sensor system to obtain the optimal transformation matrix C=C _Ie · C _sI · C _cs from the camera body coordinate system to the GPS coordinate system that is ultimately required;

Step 2: Use the phase coordinates of the corresponding connected domain and the camera field of view to calculate the coordinates of the camera optical path direction vector corresponding to the corresponding connected domain in the camera body coordinate system The preferred calculation formula is as follows:

where θ _x is the angle between the projection of the optical path direction vector on the XoZ plane and the z-axis, and the calculation formula is θ _x =(x _i -u/2)/u; similarly, θ _y is the optical path direction vector on the YoZ plane The angle between the projection of and the z-axis, the calculation formula is: θ _y =(y _i -v/2)/v;

Step 3: Use the optical path direction vector calculated in the second step and the transformation matrix calculated in the first step to obtain the coordinates of the optical path direction vector in the GPS coordinate system

Step 4: Use the results of the third step and the coordinates of the vector from the satellite centroid to the earth's center provided by the satellite in the GPS coordinate system Solving equations: Among them, R+h is the distance from the position of the moving target to the center of the earth, R is the radius of the earth at the corresponding location, h is the altitude of the target, which is a preset value. Different heights can be set by yourself according to different targets, and the unknown is the camera to the target. The distance r and the direction vector of the target to the center of the earth in the GPS coordinate system The specific practical problems are shown in Figure 3. The corresponding variables in the figure are the same as those involved in the text. Specifically, using the triangular cosine theorem formula:

(R+h) ² =r ² +l ² -2rlcos(α _i )

where α _i is the angle between the reverse vector from the camera to the target and the vector from the satellite's center of mass to the earth's center. Since the above two quantities are known quantities, they can be directly obtained. Reuse formula The vector of the direction vector from the target to the center of the earth in the GPS coordinate system can be obtained Finally, the coordinate vector of the target in the GPS coordinate system can be obtained: l _i =(R+h)·l _to .

(5) Based on the target speed and direction threshold, extract the connected domain of the real moving target, and use its GPS coordinates to calculate the geographic location of the moving target, calculate the speed and direction of the moving target, set the speed and direction threshold of the moving target, and determine The calculated effective connected domain that satisfies the threshold condition is the moving target, otherwise the connected domain that does not meet the condition is eliminated to realize the moving target detection and its speed estimation. details as follows:

Step 1: After completing the coordinate calculation of the effective connected domain centroid of each frame in the GPS coordinate system, the connected domain extracted from the current frame image is sorted according to the area and energy of the connected domain. First, sort according to the area of the connected domain, and place the largest area at the front of the queue. For connected domains with the same area, compare the sum of the gray values of the pixels and sort by size; if they are all the same, maintain the current sorting and put The coordinates, area, energy, etc. of each connected domain are stored under the respective connected domain number. (According to the size of the storage space, the information of the connected domain of N-frame images with a fixed number of frames is stored in a rolling manner, and the information of the initial frame is covered after the number of frames is exceeded, and so on.) Starting from the second frame, perform steps 2 to 5 ;

Step 2: Based on the connected domain information of the image captured by the current frame, the connectivity detection of each connected domain (current connected domain) and the previously stored connected domain information is performed frame by frame by connected domain (detected connected domain). The third step and the fourth step are respectively performed according to the length of the detected connectivity domain.

If the length of the detected connected domain is 1, perform the third step: judge only based on the motion speed between the two connected domains; if the length of the detected connected domain is greater than 1, perform the fourth step: based on the two connected domains Judging the movement speed and movement direction between domains;

Step 3: When the length of the detected connected domain is only 1, only the criterion of the motion speed between the two connected domains can determine the connectivity of the two connected domains. Using the difference between the coordinates recorded by the two in the GPS coordinate system, the displacement between the two can be obtained, and then using the frame number of the detected connected domain to make a difference with the current frame, and using the exposure interval T to obtain two frames the time interval between and calculate the corresponding motion velocities of the two connected domains. Compare the calculated movement speed with the set target movement speed threshold. (For example, the movement speed range of the aircraft is preferably 200m/s~800m/s, and a certain error tolerance is introduced, preferably set at 90%~110%, then the upper limit of the movement speed threshold of the final target is preferably 880m/s, The lower limit is preferably 180m/s) when the motion speed is not within the threshold interval, jump back to the second step to continue the loop; when the motion speed conforms to the interval, it is preferable to add 1 to the length of the current connected domain, and dump all the information of the detected connected domain. At the current position of the connected domain, delete all the contents stored in the detected connected domain, save the calculated motion speed, calculate and store the motion direction vector corresponding to the two connected domains. Jump back to the second step, change the current connected domain to the next connected domain, and enter a new loop.

Step 4: When the length of the detected connected domain is greater than 1, the connectivity of the two connected domains can be judged according to the movement speed and direction of the connected domain. Extract the connected domain information of the latest frame in the detected connected domain, and calculate the corresponding motion speed and motion direction vector together with the detected connected domain information according to the third step. Compare the calculated movement speed with the set target movement speed threshold. When the movement speed is within the threshold range: Do a dot product between the calculated movement direction vector and the movement direction vector stored in the detected connected domain, and obtain The cosine value of the angle between the two motion direction vectors is obtained and compared with the set cosine threshold value. (For example, the change of the moving direction of the moving target in a short period of time will not be very large, so it is considered that the included angle of the moving direction should preferably not exceed 20° between multiple frames, and the corresponding cosine value is preferably 0.94, that is, the cosine threshold is preferably set. is 0.94, when the dot product result of the two motion direction vectors is less than the threshold, it is considered that the current connected domain is not the target corresponding to the detected connected domain, and it is judged that the two are not connected) When the comparison result is that the two are not connected, jump back The second step continues to execute the loop; when the comparison result is that the two are connected, the length of the current connected domain becomes the length of the connected domain to be detected plus 1, and all the information of the connected domain to be detected is transferred to the location of the current connected domain and the detected connected domain is deleted. For all the contents stored in the connected domain, the calculated movement speed and movement direction are weighted based on the length of the detected connected domain and the stored movement speed and movement direction of the detected connected domain, and stored in the current connected domain. Jump back to the second step, change the current to the next connected domain, and enter a new loop. The principle of connectivity judgment is shown in Figure 4. In the figure, u _ij , the numbers of i and j respectively represent the jth connected domain in the ith frame image, and u represents the x-axis coordinate of the centroid of the corresponding connected domain on the phase plane. , v represents the y-axis coordinate of the centroid of the corresponding connected domain on the phase plane.

Step 5: After the second step is completed, examine all connected domains stored in the current frame, preferably when the length of the connected domain is greater than 2, send a detection success data packet to the onboard computer. The data includes the detection success sign, the motion speed of the detected target, its motion direction vector in the GPS coordinate system, and the position coordinates of the target in the GPS coordinate system, etc., for the onboard computer to judge.

The present invention shoots through the prototype and realizes the real-time operation of the algorithm. The results show that the method can accurately extract suspicious moving objects in the image in real time and calculate the size and direction of the moving speed of the objects.

The parts not described in detail in the present invention belong to the common knowledge of those skilled in the art.

Claims (10)

1. A satellite-borne moving target detection and speed estimation method is characterized by comprising the following steps:

(1) carrying out background suppression processing on a plurality of continuously shot satellite-borne earth observation images to obtain background-removed images;

(2) performing threshold segmentation on the background-removed image in the step (1), then performing connected domain extraction, and removing part of false target connected domains to obtain effective connected domains;

specifically, threshold segmentation is carried out on the background-removed image by using a self-adaptive threshold method, and coordinates and gray values of all pixels higher than a threshold value are stored; extracting connected domains from the stored pixels based on a four-connected domain method, and eliminating partial false target connected domains according to the area and length-width ratio threshold values of the extracted connected domains to obtain effective connected domains;

(3) carrying out centroid extraction on the effective connected domain in the step (2) and solving an accurate coordinate of the centroid of the effective connected domain on the phase plane of the shooting camera;

(4) the coordinate conversion from the accurate coordinate of the centroid on the phase plane of the shooting camera to the GPS coordinate system is completed by utilizing the coordinate conversion matrix and the camera optical axis pointing parameter, and the effective connected domain centroid coordinate under the GPS coordinate system is obtained;

(5) obtaining a geographic position corresponding to the centroid of the corresponding effective connected domain according to the centroid coordinates of the effective connected domain in the GPS coordinate system, calculating the corresponding movement speed and direction of the geographic position, extracting a real target by using a method of comparing multi-frame image connected domains, extracting the connected domain corresponding to the real moving target, finally realizing moving target detection and realizing speed estimation of the moving target.

2. The method for detecting and estimating the speed of the satellite-borne moving target according to claim 1, wherein the method comprises the following steps: the continuously shot multiframe satellite-borne earth observation images in the step (1) require that: the early warning camera installed on an orbiting satellite with an orbit height of 500km shoots, and the view field angle of the early warning camera is 4 degrees.

3. The method for detecting and estimating the speed of the satellite-borne moving target according to claim 1, wherein the method comprises the following steps: the method comprises the following steps of (1) carrying out background suppression processing on a plurality of continuously shot multi-frame satellite-borne earth observation images to obtain background-removed images, wherein the background-removed images specifically comprise the following steps: and respectively processing each frame of image, and performing background suppression to obtain a background-removed image.

4. The method for detecting and estimating the speed of the satellite-borne moving target according to claim 1, wherein the method comprises the following steps: extracting a connected domain, specifically: and (3) roughly extracting all potential target connected domains based on a four-connected domain method, calculating the area of each connected domain and the maximum length-width ratio of the connected domains, comparing the area and the maximum length-width ratio with a set fixed threshold, and eliminating false target connected domains which do not meet the requirements to obtain the remaining connected domains as effective connected domains.

5. The method for detecting and estimating the speed of the satellite-borne moving target according to claim 1, wherein the method comprises the following steps: the method comprises the following steps of converting the coordinates of a communication domain centroid phase plane into the coordinates of a GPS coordinate system, specifically: using the transformation matrix C from the camera body coordinate system to the GPS coordinate system, the satellite displacement between the time and the earth centerRepresenting in a GPS coordinate system, corresponding to the position earth radius R; according to the direction vector of the optical path direction corresponding to the accurate coordinate of the connected domain on the phase surface in the camera body coordinate systemBy the formula:

in the formulaThe direction vector of the corresponding light path of the pixel under the GPS coordinate system is taken as the direction vector of the pixel; by the formula:

in the formulaThe direction vector of the object corresponding to the connected domain relative to the earth center under the GPS coordinate system, h is the height of the target, r is the distance from the satellite to the object corresponding to the connected domain, and the expression of r and the direction vector under the GPS coordinate system is solved according to the formula

6. The method for detecting and estimating the speed of the satellite-borne moving target according to claim 1, wherein the method comprises the following steps: the method for extracting the real target by comparing the connected domains of the multi-frame images specifically comprises the following steps:

(1) sequencing the communication domains extracted from each frame of image according to the areas and the energies of the communication domains; specifically, sorting is carried out according to the area of the connected domains, the largest area is arranged at the front of the queue, and for the connected domains with the same area, the sum of pixel gray values of the connected domains is compared and sorted according to the size;

(2) and starting from the second frame, carrying out frame-by-frame connected domain matching with the connected domains in all the previous frames one by one, and storing the successfully matched connected domain combination in the corresponding connected domain of the current frame.

7. The method according to claim 6, wherein the method comprises: and sequencing the communication domains extracted from each frame of image according to the areas and the energies of the communication domains, specifically, sequencing according to the areas of the communication domains, placing the communication domains with the largest areas at the forefront of the queue, comparing the sums of the gray values of the pixels of the communication domains with the same areas, and sequencing according to the sizes.

8. The method for detecting and estimating the speed of the satellite-borne moving target according to claim 1, wherein the method comprises the following steps: the suspicious target area is accurately extracted under the condition of small calculated amount, and more accurate calculation results of speed and direction can be provided, so that the false alarm rate is reduced, and the detection success rate is improved.

9. The method for detecting and estimating the speed of the satellite-borne moving target according to claim 1, wherein the method comprises the following steps: the method is used in a satellite-borne early warning camera, and realizes on-orbit real-time moving target detection and speed estimation.

10. The method for detecting and estimating the speed of the satellite-borne moving target according to claim 1, wherein the method comprises the following steps: suspicious target information is provided to the satellite in real time.