A satellite image anchoring method and system

Abstract

This invention discloses a satellite image anchoring method and system, relating to the field of image anchoring. The satellite image anchoring method includes: scanning a map along a preset path and displaying satellite images showing changes in the current field of view frame by frame; identifying one or more targets in the center of the satellite images and storing the identified coordinate information in a list; and determining whether the target before and after movement is the same by reducing the difference in the target's deviation from the direction of movement and amplifying the difference in the target's shape change based on the coordinate information. By designing an anchoring function and utilizing the reduction of deviation direction changes and the amplification of target shape changes to design a target matching method, the aforementioned technical problems are solved.

Description

Technical Field

[0001] This invention relates to the field of image anchoring, and more specifically to a satellite image anchoring method and system. Background Technology

[0002] In deep learning-based target tracking algorithms, they are generally categorized into two frameworks based on whether the tracking module is integrated with the target detection network: detection-based tracking and joint detection-tracking. Classic detection-based tracking algorithms include Meanshift, Particle Filter, and Kalman Filter. The Meanshift algorithm, with its core algorithm, can determine the target location by obtaining a convergent optimal probability density distribution with limited computation, but it is prone to losing or mislocating targets due to changes in target scale. While Particle Filter offers greater robustness, it requires a larger number of samples to accurately describe the posterior probability distribution, significantly increasing computational demands. Furthermore, it is prone to tracking failures when the target scale changes. Correlation filtering methods are centered around filters; as filter models become more complex, the computational burden increases significantly. For example, the Kernel Correlation Filter (KCF) algorithm requires substantial computation, making it unsuitable for high real-time scenarios. It also relies heavily on cyclic matrices and performs poorly with targets at multiple scales. Joint detection and tracking frameworks require the use of feature images to calculate apparent feature vectors, significantly increasing the computational cost and reducing real-time performance. Currently, in multi-target tracking, especially in real-time scenarios, it is necessary to consider and address issues such as target loss during detection and repeated counting after a missed target is detected again. Therefore, a satellite image anchoring method and system are needed to achieve computationally efficient real-time tracking and matching of multiple targets, and to stably track moving objects with low confidence levels. Summary of the Invention

[0003] The technical problem to be solved by this invention is to achieve real-time tracking and matching of multiple targets with low computational cost and to stably track moving objects. The purpose is to provide a satellite image anchoring method and system. By designing an anchoring function and using a target matching method that reduces deviation direction changes and amplifies target shape changes, the above-mentioned technical problem is solved.

[0004] This invention is achieved through the following technical solution:

[0005] A satellite image anchoring method includes: scanning a map along a preset path and displaying satellite images showing changes in the current field of view frame by frame; identifying one or more targets in the center of the satellite images and storing the identified coordinate information in a list; and reducing the difference in the target's deviation from the direction of motion and amplifying the difference in the target's shape change based on the coordinate information, thereby determining whether the target before and after the motion is the same.

[0006] Move along a preset path and acquire satellite images of the changing field of view frame by frame, including: using the pyautogui library to perform a straight-line scan operation on the electronic map along an S-shaped path and displaying satellite images of the current field of view frame by frame.

[0007] The system scans the map along a preset path, acquiring satellite images of the changing current field of view frame by frame; it identifies one or more targets in the center of the satellite images, including: identifying targets in a preset central region of the satellite images, and identifying targets in the surrounding white space when the satellite images, excluding the preset central region, move.

[0008] Before scanning the map along a preset path and displaying satellite images showing changes in the current field of view frame by frame, the process includes: acquiring a set of map sample images; performing image enhancement processing on the set of map sample images to obtain an image enhancement dataset; calling a YOLOv5 neural network model using a weight file trained on the image enhancement dataset; and using the model to perform image enhancement processing on the satellite images to identify one or more targets in the center of the satellite images.

[0009] The identified target information is stored in a list, including: creating a scrolling box using scrolledtext.ScrolledText(), and outputting the identified target information into a text box to form a storage list.

[0010] The target information mentioned above includes any one or more of the following: class name number, center point x-axis coordinate, center point y-axis coordinate, relative width of the prior bounding box, and relative length of the prior bounding box.

[0011] The system scans the map along a preset path and displays satellite images showing changes in the current field of view frame by frame. This includes using the mss library and win32api library to capture screen pixels of the map and detect the screen pixel size in order to display the satellite images showing changes in the current field of view frame by frame.

[0012] To reduce the difference in target deviation from the direction of motion and amplify the difference in target shape changes, thereby determining whether the target before and after motion is the same, this includes: designing a characterization of the displacement correlation between targets, expressed as: The design incorporates functions representing changes in the target size, including relative size λ1 and absolute size λ2. The function expression for the relative size is as follows: The absolute size is obtained by amplifying the changes in the width and height of the two targets: In the formula, x1 and y1 are the x and y coordinates of the target in the previous frame, x2 and y2 are the x and y coordinates of the target in the current frame, w1 and h1 are the width and height of the target in the previous frame, w2 and h2 are the width and height of the target in the current frame, and k is an adjustable constant; the similarity is evaluated by calculating the discreteness of the target's shape changes. Substituting λ1 into x, we get λ1', where λ1 takes values ​​in the range [0, 10]. The difference score is calculated based on the target's deviation from the direction of movement and the similarity of the target's shape, and is expressed as follows: In the formula, (1-μ)×k1 represents the description of the target's deviation from the direction of movement, where μ is obtained by selecting formula (3) or formula (4) according to the target's current deviation direction; λ1′×k2 and λ2×k3 represent the description of the similarity of the target's shape; k1, k2, and k3 are adjustable constants used to reduce the difference in the target's deviation from the direction of movement and to amplify the difference in the target's shape changes, so as to determine whether the target before and after the movement is the same based on the difference score.

[0013] k1, k2, and k3 are distributed in a ratio of 3:1:1.

[0014] A satellite image anchoring system includes: a frame-by-frame scanning module that scans a map along a preset path and displays satellite images showing changes in the current field of view frame by frame; a target recognition module that identifies one or more targets in the center of the satellite image and stores the identified target information in a list; and a target matching module that reduces the difference in the target's deviation from the direction of movement and amplifies the difference in the target's shape changes, thereby determining whether the target before and after the movement is the same.

[0015] Compared with the prior art, the present invention has the following advantages and beneficial effects:

[0016] This invention provides a satellite image anchoring method and system. By scanning a map along a preset path and displaying satellite images showing changes in the current field of view frame by frame, it simulates a satellite transit effect through dynamic changes in the satellite map. By identifying one or more targets in the center of the satellite image and storing their coordinate information in a list, it obtains the detected targets within the area. By reducing the difference in the target's deviation from the direction of motion and amplifying the difference in the target's shape change based on the coordinate information, it determines whether the target before and after the movement is the same, matching the moving targets in consecutive frames. This improves the accuracy of target tracking, simplifies computation, and achieves stable tracking of moving objects. This invention, through the design of an anchoring function and the design of a target matching method that reduces the change in deviation direction and amplifies the change in target shape, achieves real-time tracking and matching of multiple targets with low computational complexity, enabling stable tracking of moving objects. Attached Figure Description

[0017] To more clearly illustrate the technical solutions of the exemplary embodiments of the present invention, the accompanying drawings used in the embodiments will be briefly described below. It should be understood that the following drawings only show some embodiments of the present invention and should not be considered as a limitation of the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort. In the drawings:

[0018] Figure 1 This is a schematic diagram of the satellite image anchoring method in Embodiment 1 of this application;

[0019] Figure 2 This is a schematic diagram of target detection in Embodiment 1 of this application;

[0020] Figure 3 This is a schematic diagram of the detection path in Embodiment 1 of this application;

[0021] Figure 4 This is a schematic diagram of the image enhancement processing in Embodiment 1 of this application;

[0022] Figure 5 This is a schematic diagram of the satellite image anchoring system of Embodiment 2 of this application. Detailed Implementation

[0023] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the embodiments and accompanying drawings. The illustrative embodiments and descriptions of this invention are only for explaining this invention and are not intended to limit this invention.

[0024] Example 1

[0025] like Figure 1As shown, this application provides a satellite image anchoring method, including: scanning a map along a preset path and displaying satellite images showing changes in the current field of view frame by frame; identifying one or more targets in the center of the satellite images and storing the identified coordinate information in a list; and reducing the difference in the target's deviation from the direction of movement and amplifying the difference in the target's shape change based on the coordinate information, thereby determining whether the target before and after the movement is the same.

[0026] The system simulates satellite transit by dragging an electronic map along a defined path and displaying the images frame by frame, storing the coordinates of each identified target in a list. Using this coordinate information, the system identifies target deviation directions and shape changes, minimizing the difference in target deviation direction and amplifying the difference in target shape, thus determining whether the target in each frame is the same. Moving the map and scanning frame by frame allows multiple targets to share common relative motion conditions, improving target identification accuracy. By designing an anchoring function, without employing complex neural network feature maps or computationally intensive correlation filtering techniques, the system achieves high-precision multi-target tracking of military satellite images with minimal computational cost.

[0027] Move along a preset path and acquire satellite images of the changing field of view frame by frame, including: using the pyautogui library to perform a straight-line scan operation on the electronic map along an S-shaped path and displaying satellite images of the current field of view frame by frame.

[0028] This system enables the mouse to automatically drag the map to simulate the effect of a satellite passing over it. It utilizes multi-threading to automatically drag the map, detect targets on the captured map, match the detection results using a multi-target tracking algorithm, and visualize the detection results. This achieves stable tracking and avoids flickering phenomena caused by frequent detections and non-detections.

[0029] The system scans the map along a preset path, acquiring satellite images of the changing current field of view frame by frame; it identifies one or more targets in the center of the satellite images, including: identifying targets in a preset central region of the satellite images, and identifying targets in the surrounding white space when the satellite images, excluding the preset central region, move.

[0030] like Figures 2-3 As shown, the target moves along an "S"-shaped path, and the movement of the target within the field of view in the same time period is related to the x or y axis. A judgment is added during detection: whether the target is within three-fifths of the screen's center (the preset central area). Only targets within three-fifths of the screen's center are retained. The surrounding blank areas are detected as the field of view moves downwards.

[0031] During target detection, targets may be identified but then lost due to indistinct features. To address this, a target anchoring mechanism is proposed: when a target's coordinates are within the detection area but not yet at its edge, anchoring is performed frame-by-frame, preserving these coordinates. The coordinates are then updated in the last frame based on frame count changes, continuing until the target's coordinates exceed the detection area, at which point its information is removed from the list, thus releasing the anchor. If a target is occluded or has indistinct features leading to low confidence and frequent missed detections, this anchoring mechanism ensures that information on frequently missed targets remains consistent, allowing for rapid matching once the target is detected again.

[0032] Before scanning the map along a preset path and displaying satellite images showing changes in the current field of view frame by frame, the process includes: acquiring a set of map sample images; performing image enhancement processing on the set of map sample images to obtain an image enhancement dataset; calling a YOLOv5 neural network model using a weight file trained on the image enhancement dataset; and using the model to perform image enhancement processing on the satellite images to identify one or more targets in the center of the satellite images.

[0033] like Figure 4 As shown, image enhancement processing can be achieved by rotating each map sample image counterclockwise multiple times at a preset angle and adding different numbers of markers, such as black squares, at random locations in the image. A detector built using a trained neural network model extracts satellite map data and performs target detection in the detection area. Using YOLOv5 neural network detection achieves lightweight processing.

[0034] In the application, a total of 9618 rotated and covered images were divided into training, validation, and test sets in a 7:4:3 ratio for training. The epoch was set to 100, the batch size to 5, and weight files trained using mini-batch data were used for training. The training curve shows that pre-training effectively improved the model's convergence speed. The model converged in just 95 epochs.

[0035] The identified target information is stored in a list, including: creating a scrolling box using scrolledtext.ScrolledText(), and outputting the identified target information into a text box to form a storage list.

[0036] The target information mentioned above includes any one or more of the following: class name number, center point x-axis coordinate, center point y-axis coordinate, relative width of the prior bounding box, and relative length of the prior bounding box.

[0037] The system scans the map along a preset path and displays satellite images showing changes in the current field of view frame by frame. This includes using the mss library and win32api library to capture screen pixels of the map and detect the screen pixel size in order to display the satellite images showing changes in the current field of view frame by frame.

[0038] To reduce the difference in target deviation from the direction of motion and amplify the difference in target shape changes, thereby determining whether the target before and after motion is the same, this includes: designing a characterization of the displacement correlation between targets, expressed as: The design incorporates functions representing changes in the target size, including relative size λ1 and absolute size λ2. The function expression for the relative size is as follows: The absolute size is obtained by amplifying the changes in the width and height of the two targets: In the formula, x1 and y1 are the x and y coordinates of the target in the previous frame, x2 and y2 are the x and y coordinates of the target in the current frame, w1 and h1 are the width and height of the target in the previous frame, w2 and h2 are the width and height of the target in the current frame, and k is an adjustable constant; the similarity is evaluated by calculating the discreteness of the target's shape changes. Substituting λ1 into x, we get λ1', where λ1 takes values ​​in the range [0, 10]. The difference score is calculated based on the target's deviation from the direction of movement and the similarity of the target's shape, and is expressed as follows: In the formula, (1-μ)×k1 represents the description of the target's deviation from the direction of movement, where μ is obtained by selecting formula (3) or formula (4) according to the target's current deviation direction; λ1′×k2 and λ2×λ3 represent the description of the similarity of the target's shape; k1, k2, and k3 are adjustable constants used to reduce the difference in the target's deviation from the direction of movement and to amplify the difference in the target's shape changes, so as to determine whether the target before and after the movement is the same based on the difference score.

[0039] First, an evaluation function for the center coordinates is designed, and a μ value representing the displacement correlation between targets is designed. x and μ y A function representing the change in target size is designed, including changes in relative size λ1 and absolute size λ2. To distinguish multiple targets of the same type and similar shape but different sizes in the same scene, the changes in the width and height of the prior boxes of two targets are amplified. Since the change in the prior box of the same target between two frames is small, the calculation result of λ1 for the same target in equation (5) will be very close to 1. Therefore, the closer λ1 is to 1, the higher the score of the relative size feature should be.

[0040] The difference score threshold is used to determine whether two objects are the same target. When calculating the deviation of a target from its direction of movement, the smaller the deviation, the more likely they are the same target. When describing the similarity of target shapes, a coefficient is used to amplify the differences between different targets.

[0041] Anchoring is performed frame-by-frame for one or more targets within the detection area, including: when the coordinates of a target within the detection area are not yet at the edge of the detection area, frame-by-frame anchoring preserves the target's coordinate information, and the list of coordinate information is updated based on the motion state of the last frame; when a target moves out of the detection area, the anchoring is released, including: when the target coordinates exceed the detection area, the coordinate information is deleted from the list, i.e., the anchoring is released. During frame-by-frame anchoring, the coordinate information is continuously updated based on the motion state of the last frame, and the information is deleted from the list when anchoring is released.

[0042] Optionally, k1, k2, and k3 can be distributed in a 3:1:1 ratio, which can better match the displacement changes of objects while incorporating the object's shape and size features. Optionally, matched targets need to have their ID information added and be removed from list_0. After all targets in list_0 have been matched, list_0 is checked for emptiness, i.e., it is analyzed whether any targets have disappeared. If a target with that ID is empty in list_0, it is considered a successful match; otherwise, it is considered an unsuccessful match, and the target disappears. If the target is located in the central region of the image, an anchoring mechanism is used to update its coordinate information from list_0 to list_1. List_1 is then used to store the coordinate information for the next frame.

[0043] Example 2

[0044] like Figure 5 As shown, this application provides a satellite image anchoring system, including: a frame-by-frame scanning module: scanning on a map along a preset path and displaying satellite images of the current field of view changes frame by frame; a target recognition module: recognizing one or more targets in the middle of the satellite image and storing the recognized coordinate information in a list; and a target matching module: reducing the difference in the target's deviation from the direction of movement and amplifying the difference in the target's shape changes based on the coordinate information, thereby determining whether the target before and after the movement is the same.

[0045] The principles of the embodiments in this application are the same as those in Embodiment 1, and will not be described again here.

[0046] In summary, this application proposes a satellite image anchoring method and system:

[0047] By scanning the map along a preset path and displaying satellite images showing changes in the current field of view frame by frame, the system simulates satellite transit through dynamic changes in the satellite map. It identifies one or more targets in the center of the satellite image and stores their coordinate information in a list to obtain the detected targets within the area. By reducing the difference in the target's deviation from the direction of movement and amplifying the difference in the target's shape change based on the coordinate information, it determines whether the target before and after the movement is the same, matching the moving targets in consecutive frames. This improves the accuracy of target tracking, simplifies computation, and achieves stable tracking of moving objects. This invention, through the design of an anchoring function and the use of reducing deviation direction changes and amplifying target shape changes to design a target matching method, achieves real-time multi-target tracking and matching with low computational complexity, enabling stable tracking of moving objects.

[0048] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above description is only a specific embodiment of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A satellite image anchoring method, characterized by, Comprise: Scanning on the map with a preset path, displaying the satellite image of the current field of view frame by frame; Identify one or more targets in the middle of the satellite image, and store the identified coordinate information into the list; According to the coordinate information, reduce the difference of the target deviating from the motion direction, and enlarge the difference of the target shape change, so as to judge whether the target before and after the motion is the same one; The specific process is: Design the displacement correlation between targets, expressed as: (3)； (4)； The functions characterizing the target size variation include relative size and absolute size , the function expression of which is: (5)； The absolute size is obtained by amplifying the change of the width and height of the two targets: (6)； wherein, and is the horizontal coordinate and vertical coordinate of the target at the last frame, and is the horizontal coordinate and vertical coordinate of the target existing in the current frame, and is the width and height of the target at the last frame, and is the width and height of the target in the current frame, and k is an adjustable constant. Calculate the dispersion degree of the target shape change to evaluate the similarity: (7)； Substitute x to get The value of x is in [0, 10]. , The value of x is in [0, 10]. (8)； According to the target deviation from the motion direction and the target shape similarity, the difference score is calculated, expressed as: (9)； In the formula, represents a description of the target deviating from the motion direction, wherein is selected according to the current deviating direction of the target, and formula (3) or formula (4) is obtained; and represents a description of the target shape similarity; , , is an adjustable constant, which is used to reduce the difference of the target deviating from the motion direction, and enlarge the difference of the target shape change, so as to determine whether the target before and after the motion is the same one according to the difference score.

2. A method of anchoring a satellite image according to claim 1, characterized in that, Move along a preset path, and acquire the satellite image of the current field of view frame by frame, including: the mouse performs linear scanning operation on the electronic map along the s-shaped path through the pyautogui library, and displays the satellite image of the current field of view frame by frame.

3. The method of claim 1, wherein the step of determining the anchor point comprises the steps of: determining a first anchor point; determining a second anchor point; and determining the anchor point based on the first anchor point and the second anchor point. Scanning on the map with a preset path, and acquiring the satellite image of the current field of view frame by frame; Identify one or more targets in the middle of the satellite image, including: identify the target in the preset middle area of the satellite image, and identify the target in the surrounding white space of the preset middle area when the satellite image moves.

4. The method of claim 1, wherein the step of determining the anchor point comprises the steps of: determining a first anchor point; and determining a second anchor point. Before scanning on the map with a preset path and displaying the satellite image of the current field of view frame by frame, including: collecting a set of map sample images; image enhancement processing is performed on the set of map sample images to obtain an image enhancement dataset; a YOLOv5 neural network model is called through a weight file trained by the image enhancement dataset; after image enhancement processing of the satellite image by the model, one or more targets in the middle of the satellite image are identified.

5. The method of claim 1, wherein the step of determining the anchor point is performed by a computer program. Store the identified coordinate information into the list, including: create a scroll box through scrolledtext.ScrolledText(), and output the identified coordinate information into the text box to form a storage list.

6. The method of claim 1, wherein the step of determining the anchor point comprises the steps of: determining a first anchor point; determining a second anchor point; and determining the anchor point based on the first anchor point and the second anchor point. The coordinate information includes any one or more of class name number, center point x-axis coordinate, center point y-axis coordinate, prior box relative width and prior box relative length.

7. The method of claim 1, wherein the step of determining the anchor point comprises the steps of: determining a first anchor point; determining a second anchor point; and determining the anchor point based on the first anchor point and the second anchor point. Scanning on the map with a preset path, and displaying the satellite image of the current field of view frame by frame, including: using mss library and win32api library to intercept the screen pixels of the map and detect the screen pixel size, so as to display the satellite image of the current field of view frame by frame.

8. The method of claim 1, wherein the step of determining the anchor point comprises the steps of: determining a first anchor point; determining a second anchor point; and determining the anchor point based on the first anchor point and the second anchor point. k1, k2, k3 are distributed in the ratio of 3:1:

1.

9. A satellite image anchoring system, characterized by, Comprise: Frame scanning module: scanning on the map with a preset path, displaying the satellite image of the current field of view frame by frame; Target identification module: identify one or more targets in the middle of the satellite image, and store the identified target information into the list; Target matching module: reduce the difference of the target deviating from the motion direction, and enlarge the difference of the target shape change, so as to judge whether the target before and after the motion is the same one, the specific implementation process of the target matching module is: Design the displacement correlation between targets, expressed as: (3)； (4)； The function characterizing the change in the size of the target includes a relative size and an absolute size , the function expression of the relative size being (5)； The absolute size is obtained by amplifying the change of the width and height of the two targets: (6)； wherein, and is the horizontal coordinate and the vertical coordinate of the target at the last frame, and is the horizontal coordinate and the vertical coordinate of the target existing in the current frame, and is the width and the height of the target at the last frame, and is the width and the height of the target in the current frame, and k is an adjustable constant. The discrete degree of the target shape change is calculated to evaluate the similarity: (7)； Substitute x to get The value of x is in [0, 10] , The value of x is in [0, 10] (8)； The difference score is calculated according to the target deviation direction and the target shape similarity, and is expressed as: (9)； In the formula, represents a description of the target deviating from the motion direction, wherein is selected according to the current deviating direction of the target, and formula (3) or formula (4) is obtained; and represents a description of the target shape similarity; , , is an adjustable constant, used to reduce the difference of the target deviating from the motion direction, and to enlarge the difference of the target shape change, so as to determine whether the target before and after the motion is the same one according to the difference score.

Images