A fast recognition and high-precision extraction method for autonomous navigation of dim and weak space targets

Abstract

The application discloses a fast recognition and high-precision extraction method for autonomous navigation of a space dim and weak target, which comprises the following steps: image superposition, star identification, mask processing, matching template construction, convolution processing, segmentation threshold processing, connected domain detection, target region extraction and target point position acquisition. The application realizes fast recognition and high-precision extraction of a space dim and weak target in a complex space background under the condition that the performance of a star sensor is limited.

Description

Technical Field

[0001] This invention relates to a rapid identification and high-precision extraction method for autonomous navigation of faint targets in space, belonging to the field of spacecraft overall technology. Technical Background In the aerospace field, space targets generally refer to man-made orbiting vehicles that orbit the Earth outside the atmosphere, including both operational and defunct satellites, rockets, and space debris. According to statistics from the U.S. Space Targets Monitoring Network, the total number of cataloged space targets in Earth orbit is approximately 17,800. This does not include space targets smaller than 10 cm. There are approximately 500,000 space targets between 1 and 10 cm in size, and the number of targets with a diameter less than 1 cm exceeds 100 million. Furthermore, the number of space targets continues to grow rapidly, accompanied by a strain on space resources and a gradual deterioration of the space security situation.

[0002] To continuously monitor space targets and safeguard space assets, many spacecraft have been launched abroad specifically for space target observation. However, the visible arc of a single spacecraft is short and the revisit period is long, making it unsuitable for sustained tracking. While networking multiple spacecraft can improve observation efficiency, the cost of such constellations is high. Star sensors, as attitude measurement units commonly found on spacecraft, could, through algorithm development, enable existing spacecraft to universally possess the capability for space target monitoring and cataloging by utilizing image information acquired from star sensors to measure the motion state of space targets.

[0003] Due to limitations in star sensor performance, distant space targets appear faint in images, making them difficult to quickly identify and extract against a complex stellar background, thus hindering the provision of measurement data for orbit estimation. Therefore, it is necessary to provide a method for the rapid identification and extraction of faint space targets under conditions of constrained star sensor performance, thereby enhancing spacecraft's ability to perceive space targets. Summary of the Invention

[0004] The technical problem solved by this invention is to overcome the shortcomings of the prior art and provide a rapid identification and high-precision extraction method for autonomous navigation of faint space targets. Under the condition of limited star sensor performance, it realizes rapid identification and high-precision extraction of faint space targets in complex space backgrounds, thereby improving the spacecraft's perception capability of space targets.

[0005] The objective of this invention is achieved through the following technical solutions: A rapid identification and high-precision extraction method for autonomous navigation of faint targets in space includes: 1) Obtain n images obtained through continuous sampling ; 2) Obtain n images Overlay image ; 3) Perform star chart matching, in Extracting star image coordinate set ; 4) Based on the star image coordinate set For overlay images Perform masking to remove stars and obtain the image after masking. ; 5) Obtain the normalized stripe unit matching template ; 6) Match template using normalized stripe units Images of stars after stellar masking Perform convolution processing to obtain the image after stellar masking. convolutional images ; 7) Obtain the overlay image The mean of the background noise and variance ; 8) Based on the superimposed image The mean of the background noise and variance Obtain the convolutional image The mean of the background noise and variance ; 9) Convolutional images Adaptive thresholding is performed to obtain a binary image after removing background noise. ; 10) The binary image after removing background noise Perform connected component detection to obtain multiple connected components. ;in, The total number of connected components; 11) For connected components The image is then refined so that each connected component corresponds to a line segment with a width of one pixel. Then, endpoint detection is performed on the line segments to obtain the coordinates of the two endpoints of the line segments as coarse image coordinates. 12) Based on the rough image coordinates and point target template, obtain the start and end positions of the target corresponding to each connected component, and complete the target recognition and extraction method.

[0006] Preferably, the normalized stripe unit matching template Specifically: in, Let be the dispersion radius of the point target. The length of the stripe unit template. Let the step size be the distance the point target moves. Smaller than the side length of the image pixels. These are the image coordinates in the stripe unit template.

[0007] Preferably, superimposed images The background noise follows a Gaussian distribution.

[0008] Preferably, the convolutional image The mean of the background noise and variance Specifically: , .

[0009] Preferably, the method for adaptive threshold segmentation is as follows: in, Represents a convolutional image Image location Pixel values; Binary image Image location The binary result.

[0010] Preferably, the method for obtaining the start and end positions of the target corresponding to each connected component is as follows: 121) Using coarse image coordinates, the first image Extract two regions of a certain size; 122) Using coarse image coordinates, in the last image Extract two regions of a certain size; 123) Constructing point target templates ; Using point target templates For the first image Two regions of a certain size are extracted and convolutional calculations are performed on them respectively to obtain the convolution result of each pixel in the image; the pixel corresponding to the maximum value of the convolution result is obtained as the target starting point; 124) Using point target templates For the last image Two regions of a certain size are extracted and convolutional calculations are performed on them respectively to obtain the convolution result of each pixel in the image; the pixel corresponding to the maximum value of the convolution result is obtained as the target endpoint.

[0011] Preferably, the first image Extract two regions of a certain size, including the regions corresponding to the starting and ending points of the line segment. and the region corresponding to the end endpoint ; The last image Extract two regions of a certain size, including the regions corresponding to the starting and ending points of the line segment. and the region corresponding to the end endpoint .

[0012] Preferably, the point target template Specifically: in, The coordinates of the target point template.

[0013] Compared with the prior art, the present invention has the following advantages: (1) The method of the present invention uses the star sensor commonly installed on spacecraft as the sensor for space target perception and measurement. While the star sensor is determining the attitude, the image obtained by it is used to perform space target perception and measurement, which enables ordinary spacecraft already in orbit to have the ability to perceive and measure space targets. (2) In the process of identifying dim targets, the method of the present invention avoids generating stripe templates of different lengths, thereby reducing the number of times the matching template is convolved with the image and improving the recognition speed of dim targets. (3) In the process of extracting dark targets, the present invention adopts the method of cross-comparison between single-frame images and superimposed images, which can determine the motion sequence of spatial targets on superimposed images, obtain accurate image coordinates of stripe endpoints, and improve the extraction accuracy of the centroid of dark targets. Attached Figure Description

[0014] Figure 1 This is a flowchart of the method of the present invention; Figure 2(a) shows the initial time image acquired by the star sensor between 0 and 2 seconds in one embodiment of the present invention. ; Figure 2(b) shows the end time image acquired by the star sensor between 0 and 2 seconds in one embodiment of the present invention. ; Figure 2(c) is a superimposed image acquired by the star sensor between 0 and 2 seconds in one embodiment of the present invention. ; Figure 3 This is a diagram showing the result of identifying the target region of a dark weak point in one embodiment of the present invention; Figure 4 This is an image showing the result of extracting dark weak point targets in one embodiment of the present invention. Detailed Implementation

[0015] like Figure 1 As shown, the present invention provides a method for rapid identification and extraction of faint targets in space based on a star sensor, comprising the following steps: (1) The original image sequence of the star sensor is superimposed, and the exposure time of the star sensor is set to... Star sensor in ~ A total of [number] samples were collected during the time period. n Images, each image is represented as .in, Corresponding time The captured images, Corresponding time The captured images. ~ The overlay image of the original images acquired over a time period is represented as follows: Store using memory Three images. Because stars and space objects move differently, when a star sensor observes a star in a specific orientation, the star appears as a dot in the superimposed image, while the space object appears as stripes. The superimposed image can distinguish the space object from the star in terms of shape.

[0016] (2) Perform star identification and star catalog matching on the original images of a single star sensor. Extracting star image coordinate set (3) Based on the star image coordinate set identified in step (2), the superimposed image in step (1) is processed. Perform stellar masking to remove superimposed images The stars in the image eliminate interference from stars in the identification of faint targets in space. Centered on a radius of r, the image is processed using a mask; the image after the stellar mask is obtained. Represented as: in, , These are the image coordinates.

[0017] (4) Construct a stripe unit matching template for spatial targets. Represented as: in, The length of the stripe unit template. The direction of the stripe unit template, , Let the step size be the distance the point target moves. Smaller than the side length of a pixel. The radius of dispersion for a point target is given; this value is related to the optical system parameters and can be taken as [value missing]. , These are the image coordinates in the stripe unit template. It can be Divisible by.

[0018] Therefore, the size of the stripe unit matching template is The stripe unit matching template is normalized. Represented as: (5) Using the normalized fringe unit matching template established in step (4), perform traversal convolution calculation on the superimposed image in step (3) to obtain the image after the stellar mask. convolutional images , represented as: in, This is the convolution operator.

[0019] (6) Assuming superimposed images The background noise follows a Gaussian distribution. Represented as: in, To overlay images The mean of the background noise, To overlay images Variance of background noise This represents a Gaussian distribution.

[0020] (7) Obtain the intensity distribution after convolving the stripe unit matching template with the background noise. The convolved image after convolving the stripe unit matching template with the background. Background noise distribution Represented as: in, This represents the mean of the background noise after convolution. , This represents the variance of the background noise after convolution. .

[0021] (8) Perform adaptive thresholding on the image after convolution calculation in step (5). Let the adaptive threshold be set to... It can remove 99.73% of the background signal, resulting in a binary image after background noise removal. for; in, Represents a convolutional image Image location Pixel values; For each image location The binary result.

[0022] (9) The binary image obtained in step (8) Connected component detection is performed, where each connected component represents an image region containing a spatial object. The connected component is represented as... ,in The number of connected components, i.e., the number of spatial targets. (10) For the connected components obtained in step (9) After refining, a single-pixel-width line segment is obtained for each connected component. Then, endpoint detection is performed on the lines to obtain the approximate image coordinates of the two endpoints of the striped spatial target. In the image coordinates, the superscript 's' represents the starting endpoint (the end of the line segment closest to the origin is taken as the starting endpoint), the superscript 'e' represents the ending endpoint, the connected components are numbered, and the subscript 'i' represents the number of the connected components.

[0023] (11) Using the rough coordinate positions obtained in step (10) as the center, the single image obtained in step (1) Extract The region of size m=20 Extracting region and Represented as: (12) Using the rough coordinate positions obtained in step (10) as the center, the single image obtained in step (1) Extract The region of size m=20 Extracting region and Represented as (13) Constructing point target templates Target template The size is Target template Represented as in, The coordinates of the target point template.

[0024] (14) Use the point target template S to process the data obtained in step (11). and Perform convolution calculations on each pixel separately, and obtain the convolution result of each pixel as the output response; obtain and The image coordinates corresponding to the maximum output response in the region. yes The image coordinates I1 of the mid-space target have two regions, and the output is one maximum point. Represented as (15) Use the point target template S to process the data obtained in step (12). and Perform convolution calculations and obtain the results respectively. and The image coordinates corresponding to the maximum output response in the region. yes The image coordinates of a mid-space target are divided into two regions and one coordinate. Represented as (16) Repeat steps (10) to (15) to transform all connected components in the image. The process involves identifying and extracting all spatial targets.

[0025] Based on the above steps, a set of simulation examples are given, where the exposure time of the star sensor is... The star sensor acquired 10 images within a 0-2s time period, with each image having a noise level of [missing information]. Initial time image End time image Overlay images As shown in Figure 2.

[0026] Let the stripe unit template length be , The step size for moving the point target, and the direction of the striped unit template. Step size is The dispersion radius of the point target By using stripe unit templates to process the superimposed images, the image region where the target is located is identified, such as... Figure 3 As shown. Target extraction is performed using the identified region, and the extracted target image location is compared with the real location, for example... Figure 4 As shown, ☆ marks the extracted location of the target, ○ marks the actual location of the target, and the extraction error is [-1.85, 0.47].

[0027] The method of this invention utilizes star sensors commonly installed on spacecraft to perform space target perception and measurement. By processing the images acquired by the original star sensors, the angular measurement information of the space targets is obtained. This method enables the sensing and measurement of space targets while the star sensors are working normally.

[0028] The method of this invention utilizes the method of stacking sequential images to enhance the signal strength of faint targets in space, and constructs a matching unit template for faint targets in space based on the different motion characteristics of stars and space targets in the image coordinate system. This reduces the number of convolutions between the matching template and the image, thereby improving the recognition speed of faint targets.

[0029] The method of this invention uses a cross-comparison of single-frame images and superimposed images to determine the motion direction of spatial targets on the superimposed images and obtain the stripe endpoints accurately, thereby improving the extraction accuracy of the centroid of dark targets.

[0030] While the present invention has been disclosed above with reference to preferred embodiments, it is not intended to limit the invention. Any person skilled in the art can make possible variations and modifications to the technical solutions of the present invention using the disclosed methods and techniques without departing from the spirit and scope of the invention. Therefore, any simple modifications, equivalent changes, and alterations made to the above embodiments based on the technical essence of the present invention, without departing from the content of the technical solutions of the present invention, shall fall within the protection scope of the present invention. Where there is no conflict, the embodiments of this application and the technical features thereof can be combined with each other.

[0031] The contents not described in detail in this specification are common knowledge to those skilled in the art.

Claims

1. A rapid identification and high-precision extraction method for autonomous navigation of targets in dimly lit space, characterized in that, include: 1) Obtain n images obtained through continuous sampling ; 2) Obtain n images Overlay image ; 3) Perform star chart matching, in Extracting star image coordinate set ; 4) Based on the star image coordinate set For overlay images Perform masking to remove stars and obtain the image after masking. ; 5) Obtain the normalized stripe unit matching template ; 6) Match template using normalized stripe units Images of stars after stellar masking Perform convolution processing to obtain the image after stellar masking. convolutional images ; 7) Obtain the overlay image The mean of the background noise and variance ; 8) Based on the superimposed image The mean of the background noise and variance Obtain the convolutional image The mean of the background noise and variance ; 9) Convolutional images Adaptive thresholding is performed to obtain a binary image after removing background noise. ; 10) The binary image after removing background noise Perform connected component detection to obtain multiple connected components. ;in, The total number of connected components; 11) For connected components The image is then refined so that each connected component corresponds to a line segment with a width of one pixel. Then, endpoint detection is performed on the line segments to obtain the coordinates of the two endpoints of the line segments as coarse image coordinates. 12) Based on the rough image coordinates and point target template, obtain the start and end positions of the target corresponding to each connected component, and complete the target recognition and extraction method; The normalized stripe unit matching template Specifically: in, Let be the dispersion radius of the point target. The length of the stripe unit template. Let the step size be the distance the point target moves. Smaller than the side length of the image pixels. The image coordinates are within the stripe unit template; The direction of the stripe unit template, ; The method for obtaining the start and end positions of the target corresponding to each connected component is as follows: 121) Using coarse image coordinates, the first image Extract two regions of a certain size; 122) Using coarse image coordinates, in the last image Extract two regions of a certain size; 123) Constructing point target templates ; Using point target templates For the first image Two regions of a certain size are extracted and convolutional calculations are performed on them respectively to obtain the convolution result of each pixel in the image; the pixel corresponding to the maximum value of the convolution result is obtained as the target starting point; 124) Using point target templates For the last image Two regions of a certain size are extracted and convolutional calculations are performed on them respectively to obtain the convolution result of each pixel in the image; the pixel corresponding to the maximum value of the convolution result is obtained as the target endpoint.

2. The method for rapid identification and high-precision extraction of faint targets in space for autonomous navigation according to claim 1, characterized in that, Overlay images The background noise follows a Gaussian distribution.

3. The method for rapid identification and high-precision extraction of faint targets in space for autonomous navigation according to claim 2, characterized in that, The convolutional image The mean of the background noise and variance Specifically: ， 。 4. The method for rapid identification and high-precision extraction of faint targets in space for autonomous navigation according to claim 3, characterized in that, The method for adaptive threshold segmentation is as follows: in, Represents a convolutional image Image location Pixel values; Binary image Image location The binary result.

5. A rapid identification and high-precision extraction method for autonomous navigation of faint targets in space according to any one of claims 1-4, characterized in that, First image Extract two regions of a certain size, including the regions corresponding to the starting and ending points of the line segment. and the region corresponding to the end endpoint ; The last image Extract two regions of a certain size, including the regions corresponding to the starting and ending points of the line segment. and the region corresponding to the end endpoint .

6. A rapid identification and high-precision extraction method for autonomous navigation of faint targets in space according to any one of claims 1-4, characterized in that, The point target template Specifically: in, The coordinates of the target point template.

Images