Small sky mass center extraction method based on joint pixel distribution frequency template

Abstract

The invention discloses a small sky mass center extraction method based on a joint pixel distribution frequency template, and belongs to the technical field of aerospace. The implementation method comprises the following steps of: converting an RGB noise optical image of a small celestial body into a grayscale image, and segmenting the grayscale image by taking the maximum grayscale value of all pixels in the first k columns of the grayscale image as a segmentation threshold to obtain a segmented image; marking, combining and filling the connected domains to determine a small celestial body connected image and a small celestial body connected domain, performing logical judgment on a gray image by using the connected image to obtain a small celestial body connected gray image reflecting brightness information, constructing a row and column direction pixel distribution frequency template, and performing row and column distribution frequency multiplication and smoothing on corresponding pixel points to obtain a small celestial body connected gray image; a combined pixel distribution frequency template reflecting shape information is obtained, a new representation fusing the brightness and shape information of the small celestial body is established, the centroid coordinates of the small celestial body are obtained through weighted average calculation, and centroid extraction of the small celestial body is achieved. The method has the advantages of being high in solving precision, high in solving stability, high in expansion capacity and the like.

Description

Technical Field

[0001] The invention relates to a method for extracting the centroid of a small celestial body based on a joint pixel distribution frequency template, and belongs to the technical field of aerospace. Background Art

[0002] With the continuous development of aerospace technology, deep space exploration missions targeting small celestial bodies have gradually become the focus of attention at home and abroad. Spacecraft navigation, guidance and control are the basis for the smooth implementation of related tasks. At present, small celestial body relative orbit determination based on optical measurement is widely used in small celestial body exploration due to its strong autonomy and high precision. Target centroid extraction is the premise and basis for implementing optical relative orbit determination. Generally, the optical center or centroid is used instead of the centroid to provide the necessary relative line of sight information. Deep space small celestial body targets have different characteristics from general near-Earth targets or spacecraft targets, which are mainly dim and irregular, and there are many image noise interferences. Therefore, it is very important to study the centroid extraction method for small celestial bodies.

[0003] Among the methods that have been developed for extracting target celestial bodies by using the optical center or centroid instead of the centroid, the prior art [1] (see: Jia H, Liu QY, Xie YF, et al. High Precision Centroid Extraction Algorithm of Space Target for Autonomous Optical Navigation [C] / / 2022 41st Chinese Control Conference (CCC). IEEE, 2022: 6222-6227.) proposed an improved grayscale centroid extraction algorithm to achieve sub-pixel centroid extraction, and used the regional centroid extraction method to achieve preliminary centroid extraction, and then performed a sub-pixel correction method based on the grayscale value and image gradient to obtain a more accurate centroid position.

[0004] Prior art [2] (see: Cui Pingyuan, Jia He, Zhu Shengying. Irregularity of small celestial bodies and centroid extraction and application of optical navigation [J]. Journal of Astronautics, 2021, 42(01): 83-91.) A centroid extraction method based on the minimum circumscribed figure is proposed for the case where the shape of small celestial bodies is relatively irregular. Based on edge detection of the projected image, the minimum circumscribed figure of the convex hull vertices is calculated to accurately extract the centroid of irregular small celestial bodies. This method is less affected by the spin of the small celestial body and can improve the extraction accuracy of the centroid of the small celestial body, thereby improving the accuracy of autonomous optical navigation.

[0005] At present, the two alternative forms of optical center and shape center have limitations in terms of applicable scenarios such as distance and features, and the single evaluation of the brightness or shape of the target in the image is not comprehensive enough to be regarded as the target center of mass. Through a specially designed image preprocessing method and the fusion of the brightness and shape information of small celestial bodies in noisy optical images, the accuracy of small celestial body center of mass extraction can be effectively improved. Summary of the invention

[0006] The purpose of the present invention is to provide a method for extracting the centroid of a small celestial body based on a joint pixel distribution frequency template. Aiming at the problem of extracting the centroid of a small celestial body, a segmentation threshold is designed to segment the grayscale image of the small celestial body, and a connected domain is marked and combined to determine the connected image of the small celestial body and the connected domain of the small celestial body. The connected image is used to perform logical judgment on the grayscale image to obtain a connected grayscale image reflecting brightness information, and further on the basis of the connected domain, a row and column direction pixel distribution frequency template is constructed, and the row and column distribution frequencies of the corresponding pixel points are multiplied and smoothed to obtain a joint pixel distribution frequency template reflecting shape information, and a new representation that integrates the brightness and shape information of the small celestial body is established, thereby calculating the centroid coordinates of the small celestial body and realizing high-precision centroid extraction of the small celestial body. The present invention has the advantages of high solution accuracy, strong solution stability, and strong expansion capability.

[0007] The objective of the present invention is achieved through the following technical solutions:

[0008] The invention discloses a method for extracting the centroid of a small celestial body based on a joint pixel distribution frequency template, which is an improved centroid method that integrates brightness and shape information. In order to solve the problem of high-precision centroid extraction of small celestial body targets, the RGB noise optical image of the small celestial body is converted into a grayscale image, and the grayscale image is segmented using the maximum grayscale value of all pixels in the first k columns of the grayscale image as the segmentation threshold T, distinguishing the target area greater than the segmentation threshold and the background area less than or equal to the segmentation threshold, and obtaining a segmented image; for the segmented image, select the 8-connectivity criterion to mark the connected domains belonging to the same object in the segmented image, and based on the number of pixels in the connected domain, the judgment threshold n is used. TFurther distinguish the background point area and one or more small celestial body illumination areas from the target area; combine all the small celestial body illumination areas to obtain the binary representation of the small celestial body combined connected image, further fill the local 0-value holes therein, and determine the final small celestial body connected image and the small celestial body connected domain; use the small celestial body connected image to perform logical judgment on the grayscale image to obtain the small celestial body connected grayscale image used to reflect the brightness information of the small celestial body, which is used for the subsequent small celestial body centroid extraction; on the basis of the small celestial body connected domain, construct the pixel distribution frequency template of the small celestial body in both the row and column directions; comprehensively consider the two directions of the row and column, and obtain the joint pixel distribution frequency template used to reflect the shape information of the small celestial body by multiplying and smoothing the row and column distribution frequencies of the corresponding pixels; based on the grayscale value of the small celestial body connected grayscale image, establish a new representation that integrates the brightness and shape information of the small celestial body, and obtain the centroid coordinates (u c ,v c ). The present invention can be used in multiple mission scenarios such as small celestial body detection, approach navigation guidance control, etc.

[0009] The present invention discloses a method for extracting the centroid of a small celestial body based on a joint pixel distribution frequency template, comprising the following steps:

[0010] Step 1: Convert the RGB noise optical image of the small celestial body into a grayscale image, and segment the grayscale image using the maximum grayscale value of all pixels in the first k columns of the grayscale image as the segmentation threshold T, distinguish the target area greater than the segmentation threshold and the background area less than or equal to the segmentation threshold, and obtain the segmented image to achieve threshold segmentation of small celestial bodies.

[0011] The maximum gray value of all pixels in the first k columns of the small celestial body gray image is used as the segmentation threshold T:

[0012]

[0013] Among them, g(u,v) is the grayscale value of the grayscale image, and H is the image height;

[0014] The grayscale image is segmented by the segmentation threshold T to obtain the segmented image. The grayscale value s(u,v) of the segmented image is:

[0015]

[0016] Among them, the area with non-zero gray value is the target area, and the area with zero gray value is the background area.

[0017] Step 2: For the segmented image obtained in step 1, select the 8-connectivity criterion to mark the connected domains belonging to the same object in the segmented image, and based on the number of pixels in the connected domain, determine the threshold n. TFurther distinguish the background point area and one or more small celestial body illumination areas from the target area in step 1; combine all the small celestial body illumination areas obtained in step 2.1 to obtain a binary representation of the small celestial body combined connected image, further fill the local 0-value holes therein, and determine the final small celestial body connected image and the small celestial body connected domain C; use the small celestial body connected image to perform logical judgment on the grayscale image in step 1, and obtain the grayscale value I used to reflect the brightness information of the small celestial body c The connected grayscale image of small celestial bodies (u, v) is used for the subsequent extraction of the centroid of small celestial bodies and the marking of the connected domain of small celestial bodies.

[0018] Step 2.1: For the segmented image obtained in step 1, select the 8-connectivity criterion to mark the connected domains belonging to the same object in the segmented image, and based on the number of pixels in the connected domain, determine the threshold n. T Further distinguishing the background point area and one or more small celestial body illumination areas from the target area in step 1;

[0019] The connected image with the label value l(u,v) is obtained by the 8-connectivity criterion. The regions with the same label value are the connected domains belonging to the same object. The connected domain with the largest number of known pixels belongs to the background region obtained in step 1, and the other connected domains belong to the target region. The target region contains background point regions that may only occupy a few pixels, as well as one or more small celestial body illumination regions with a moderate number of pixels.

[0020] Therefore, for different connected domains belonging to the target area, the number of pixels in each connected domain is n, and the judgment threshold n based on the number of pixels is set T , distinguish the background point area less than or equal to the judgment threshold and the small celestial body illumination area A greater than the judgment threshold from the target area.

[0021] Step 2.2: Combine all the small body illumination areas obtained in step 2.1 to obtain a binary representation of the small body combined connected image, further fill the local 0-value holes therein, and determine the final small body connected image and the small body connected domain C.

[0022] By combining the illumination areas of all small celestial bodies, we can obtain a combined connected image of small celestial bodies represented by the binary value b(u,v):

[0023]

[0024] Further fill zero to multiple possible 0-value holes, that is, reassign all 0 values ​​in all holes to 1, and determine the final small celestial body connected image represented by the binary value c(u,v), and the small celestial body connected domain C corresponding to the value 1.

[0025] Step 2.3: Use the small celestial body connected image in step 2.2 to perform logical judgment on the grayscale image in step 1, and obtain the grayscale value I used to reflect the brightness information of the small celestial body. c The connected grayscale image of the small celestial body (u, v) is used for the subsequent centroid extraction of the small celestial body.

[0026] Step 3. Based on the connected domain of the small celestial body obtained in step 2, construct the pixel distribution frequency template of the small celestial body in both row and column directions; comprehensively consider the two directions of row and column, multiply and smooth the row and column distribution frequencies of corresponding pixels, and obtain the joint pixel distribution frequency template used to reflect the shape information of the small celestial body.

[0027] Step 3.1: Based on the small celestial body connected domain obtained in step 2.2, the pixel distribution frequency templates of the small celestial body rows and columns in two directions are constructed, which are:

[0028]

[0029] Among them, N is the number of all valid pixels in the connected domain of the small celestial body, U and V are the sets of all pixels in the corresponding u-th row and v-th column respectively;

[0030] Step 3.2: In both row and column directions, multiply and smooth the row and column distribution frequencies of corresponding pixels to obtain a joint pixel distribution frequency template for reflecting the shape information of small celestial bodies:

[0031]

[0032] Among them, the Hadamard product symbol ⊙ represents element-wise multiplication;

[0033] Step 4: Based on the grayscale value of the connected grayscale image of the small celestial body obtained in step 2, a new representation is established that integrates the brightness and shape information of the small celestial body. c (u,v)·Φ(u,v), the center of mass coordinates of the small celestial body (u c ,v c ), realizing the fusion representation centroid extraction.

[0034] According to formula (6), the mass center coordinates of the small celestial body (u c ,v c ).

[0035]

[0036] It also includes step five, using the small celestial body threshold segmentation in step one, the small celestial body connected domain labeling in step two, the joint pixel distribution frequency template in step three, and the fusion representation centroid extraction method in step four to achieve high-precision small celestial body centroid extraction, providing basic information for subsequent small celestial body detection and approach navigation guidance control.

[0037] Beneficial effects:

[0038] 1. The present invention discloses a method for extracting the centroid of small celestial bodies based on a joint pixel distribution frequency template. The method specifically designs segmentation thresholds and small celestial body noise optical image preprocessing methods such as connected domain combination and filling, and constructs row and column directions and their joint pixel distribution frequency templates to fuse the brightness and shape information of small celestial bodies. The centroid extraction is further realized by weighted averaging of the fused representation, thereby improving the extraction accuracy of the centroid of small celestial bodies.

[0039] 2. The present invention discloses a method for extracting the centroid of small celestial bodies based on a joint pixel distribution frequency template. Through image preprocessing, the connected domain of small celestial bodies can be determined more accurately. The fused representation contains comprehensive information on the brightness and shape of small celestial bodies, and the solution is more stable.

[0040] 3. The present invention discloses a method for extracting the centroid of small celestial bodies based on a joint pixel distribution frequency template. The imaging characteristics of small deep-space celestial body targets can also be extended to similar long-distance optical observations of other large celestial bodies. It is suitable for centroid extraction task scenarios of a variety of different targets and has stronger expansion capabilities. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] Figure 1 It is a flow chart of a method for extracting the centroid of a small celestial body based on a joint pixel distribution frequency template of the present invention;

[0042] Figure 2 It is the three-dimensional model of the small celestial body 433Eros selected in this embodiment;

[0043] Figure 3 is a local magnified image of the noise optical image of a small celestial body in this embodiment;

[0044] FIG. 4 is a partially enlarged binary image of the segmented image and the connected image in this embodiment. Figure 4a ) is a local enlarged binary image of the segmented image in this embodiment, Figure 4b ) is a local enlarged binary image of the connected image in this embodiment;

[0045] Figure 5 is a partially enlarged image of the small celestial body mass center extraction result in this embodiment, wherein the white one is the real mass center of the small celestial body, and the red one is the mass center of the small celestial body extracted by the method of the present invention;

[0046] Figure 6 is the centroid extraction error of the small celestial body when occupying different numbers of pixels in this embodiment. DETAILED DESCRIPTION

[0047] In order to better illustrate the purpose and advantages of the present invention, the present invention is explained in detail below by performing a simulation analysis on a method for extracting the centroid of a small celestial body based on a joint pixel distribution frequency template.

[0048] This embodiment is as follows Figure 2 Taking the irregular shaped small celestial body 433Eros as an example, the center of mass coordinates are extracted.

[0049] like Figure 1 As shown, the present embodiment discloses a method for extracting the centroid of a small celestial body based on a joint pixel distribution frequency template, and the specific implementation steps are as follows:

[0050] Step 1: A threshold segmentation method is used to eliminate the influence of noise in view of the low signal-to-noise ratio of the noisy optical image of a small celestial body. The RGB noisy optical image of the small celestial body is converted into a grayscale image. The grayscale image is segmented using the maximum grayscale value of all pixels in the first k columns of the grayscale image as the segmentation threshold T. The target area greater than the segmentation threshold and the background area less than or equal to the segmentation threshold are distinguished, and a segmented image is obtained.

[0051] Considering that small celestial bodies are usually located near the center of the image during tracking observation and the non-zero grayscale value of the deep space background is mainly imaging Gaussian noise, the maximum grayscale value of all pixels in the first k columns of the small celestial body grayscale image is taken as the segmentation threshold T:

[0052]

[0053] Among them, g(u,v) is the grayscale value of the grayscale image, and H is the image height; this segmentation threshold can effectively capture the maximum noise in the background, thereby segmenting most of the background; as a preferred method, k=W / 5, that is, the first 1 / 5 of the image width W;

[0054] The grayscale image is segmented by the segmentation threshold T to obtain the segmented image. The grayscale value s(u,v) of the segmented image is:

[0055]

[0056] Among them, the area with non-zero gray value is the target area, and the area with zero gray value is the background area.

[0057] Step 2: For the segmented image obtained in step 1, select the 8-connectivity criterion to mark the connected domains belonging to the same object in the segmented image, and based on the number of pixels in the connected domain, determine the threshold n. TFurther distinguish the background point area and one or more small celestial body illumination areas from the target area in step 1; combine all the small celestial body illumination areas obtained in step 2.1 to obtain a binary representation of the small celestial body combined connected image, further fill the local 0-value holes therein, and determine the final small celestial body connected image and the small celestial body connected domain C; use the small celestial body connected image to perform logical judgment on the grayscale image in step 1, and obtain a grayscale value I that can reflect the brightness information of the small celestial body c (u,v) connected grayscale image of small celestial bodies, which is used for subsequent centroid extraction of small celestial bodies;

[0058] Step 2.1: For the segmented image obtained in step 1, select the 8-connectivity criterion to mark the connected domains belonging to the same object in the segmented image, and based on the number of pixels in the connected domain, determine the threshold n. T Further distinguishing the background point area and one or more small celestial body illumination areas from the target area in step 1;

[0059] By considering the 8-connectivity criterion of edge contact and corner contact of pixels at the same time, a connected image with a label value of l(u,v) is obtained. The regions with the same label value are connected domains belonging to the same object. The connected domain with the largest number of known pixels belongs to the background region obtained in step 1, and the other connected domains belong to the target region. The target region contains background point regions that may only occupy a few pixels, as well as one or more small celestial body illumination regions with a moderate number of pixels.

[0060] Therefore, for different connected domains belonging to the target area, assuming that the number of pixels in each connected domain is n, a judgment threshold n based on the number of pixels is set. T , distinguish the background point area less than or equal to the judgment threshold and the small celestial body illumination area A greater than the judgment threshold from the target area; as a preferred option, take n T =2;

[0061] Step 2.2, combine all the small celestial body illumination areas obtained in step 2.1 to obtain a binary representation of the small celestial body combined connected image, further fill the local 0-value holes therein, and determine the final small celestial body connected image and the small celestial body connected domain C;

[0062] Considering the complex lighting conditions in space and the irregular shapes of small celestial bodies, large shadows on the surface of small celestial bodies may separate the illumination areas of multiple small celestial bodies. Therefore, all the illumination areas of small celestial bodies are combined to obtain a combined connected image of small celestial bodies represented by the binary value b(u,v):

[0063]

[0064] In addition, considering that the rugged terrain such as the crater of a small celestial body may also cause a local shadow to appear in the middle of the illuminated area of ​​the small celestial body, and the current hypothesis of the origin of small celestial bodies does not support the appearance of holes, we further fill zero to multiple possible 0-value holes, that is, reassign all 0 values ​​in all holes to 1, and determine the final small celestial body connected image represented by the binary value c(u,v), and the small celestial body connected domain C corresponding to the value 1;

[0065] Step 2.3: Use the small celestial body connected image in step 2.2 to perform logical judgment on the grayscale image in step 1, and obtain the grayscale value that can reflect the brightness information of the small celestial body as I c (u,v) connected grayscale image of small celestial bodies, which is used for subsequent centroid extraction of small celestial bodies;

[0066] Step 3: Based on the connected domain of the small celestial body obtained in step 2.2, construct pixel distribution frequency templates in both row and column directions of the small celestial body, where the pixel distribution frequency templates in the row direction have the same value in the same row, reflecting the pixel proportion of the target shape in each row, and the same applies to the column direction; comprehensively consider both row and column directions, and obtain a joint pixel distribution frequency template that can reflect the shape information of the small celestial body by multiplying and smoothing the row and column distribution frequencies of the corresponding pixel points;

[0067] Step 3.1: Based on the small celestial body connected domain obtained in step 2.2, the pixel distribution frequency templates of the small celestial body rows and columns in two directions are constructed, which are:

[0068]

[0069] Among them, N is the number of all valid pixels in the connected domain of the small celestial body, U and V are the sets of all pixels in the corresponding u-th row and v-th column respectively; the pixel distribution frequency template in the row direction has the same value in the same row, reflecting the pixel proportion of the target shape in each row, and the same is true in the column direction;

[0070] Step 3.2: In both row and column directions, multiply and smooth the row and column distribution frequencies of corresponding pixels to obtain a joint pixel distribution frequency template that can reflect the shape information of small celestial bodies:

[0071]

[0072] Among them, the Hadamard product symbol ⊙ represents element-wise multiplication;

[0073] Step 4: Based on the grayscale value of the connected grayscale image of the small celestial body obtained in step 2.3, a new representation is established that integrates the brightness and shape information of the small celestial body. c (u,v)·Φ(u,v), the center of mass coordinates of the small celestial body (u c ,v c ):

[0074]

[0075] Step five, using the small body threshold segmentation in step one, the small body connected domain labeling in step two, the joint pixel distribution frequency template in step three, and the fusion representation centroid extraction method in step four, high-precision small body centroid extraction is achieved, providing basic information for subsequent small body detection and approach navigation guidance control.

[0076] In order to verify the feasibility and effectiveness of the method, this embodiment analyzes the simulation results.

[0077] The local enlarged image of the area where the asteroid 433Eros is located is captured from the noisy optical image. Figure 3 As shown in Figure 4, the small celestial body occupies about 170 pixels. The local enlarged binary image of the small celestial body segmentation image and the connected image is shown in Figure 4. Figure 4a ) Perform binary processing on the grayscale segmented image, where white is the target area and black is the background area. It can be seen that the target area contains a small celestial body illumination area with a large number of pixels, and the background point area that only occupies a few pixels, and the two are not connected; Figure 4b ) is the connected domain of the small celestial body in the connected image. It can be seen that the imaging part of the corresponding small celestial body is successfully determined. Using the fusion representation based on the grayscale value of the connected grayscale image of the small celestial body and the joint pixel distribution frequency template, the extraction result of the centroid coordinates of the small celestial body is as follows: Figure 5 As shown in the figure, the white cross is the real mass center of the small celestial body, and the red cross is the extracted mass center of the small celestial body. The two-dimensional mass center extraction error is only about 0.0292pixel, achieving high-precision mass center extraction at the sub-pixel level for small celestial bodies. For five noisy optical images with different numbers of pixels of small celestial bodies ranging from 50 to 250, the mass center extraction error curve of the small celestial body is shown in the figure. Figure 6 As shown, the two-dimensional centroid extraction error is stabilized below 0.09 pixel. In addition, this embodiment takes less than 300 ms, has efficient real-time solving capability, and can be applied to complex real-time task scenarios.

[0078] The specific description above further illustrates in detail the purpose, technical solutions and beneficial effects of the invention. It should be understood that the above is only a specific embodiment of the present invention, which is used to explain 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 in the scope of protection of the present invention.

Claims

1. A method for extracting the centroid of a small celestial body based on a joint pixel distribution frequency template, characterized in that: The following steps are included: Step 1: Convert the RGB noise optical image of the small celestial body into a grayscale image, and use the maximum grayscale value of all pixels in the first k columns of the grayscale image as the segmentation threshold T to segment the grayscale image, distinguish the target area greater than the segmentation threshold and the background area less than or equal to the segmentation threshold, and obtain the segmented image to achieve the threshold segmentation of the small celestial body; Step 2: For the segmented image obtained in step 1, select the 8-connectivity criterion to mark the connected domains belonging to the same object in the segmented image, and based on the number of pixels in the connected domain, determine the threshold n. T Further distinguish the background point area and one or more small celestial body illumination areas from the target area in step 1; combine all the small celestial body illumination areas obtained to obtain a binary representation of the small celestial body combined connected image, further fill the local 0-value holes therein, and determine the final small celestial body connected image and the small celestial body connected domain C; use the small celestial body connected image to perform logical judgment on the grayscale image in step 1, and obtain a grayscale value I used to reflect the brightness information of the small celestial body c (u,v) small celestial body connected grayscale image, which is used for subsequent small celestial body centroid extraction and small celestial body connected domain labeling; Step 3: Based on the connected domain of the small celestial body obtained in step 2, a pixel distribution frequency template of the small celestial body in both row and column directions is constructed; by integrating the row and column directions, the row and column distribution frequencies of the corresponding pixels are multiplied and smoothed to obtain a joint pixel distribution frequency template for reflecting the shape information of the small celestial body; Step 4: Based on the grayscale value of the connected grayscale image of the small celestial body obtained in step 2, a new representation is established that integrates the brightness and shape information of the small celestial body. c (u,v)·Φ(u,v), the center of mass coordinates of the small celestial body (u c ,v c ), realizing the fusion representation centroid extraction.

2. A method for extracting the centroid of a small celestial body based on a joint pixel distribution frequency template as claimed in claim 1, characterized in that: It also includes step five, using the small celestial body threshold segmentation in step one, the small celestial body connected domain labeling in step two, the joint pixel distribution frequency template in step three, and the fusion representation centroid extraction method in step four to achieve high-precision small celestial body centroid extraction, providing basic information for subsequent small celestial body detection and approach navigation guidance control.

3. A method for extracting the centroid of a small celestial body based on a joint pixel distribution frequency template as claimed in claim 1 or 2, characterized in that: In step one, The maximum gray value of all pixels in the first k columns of the small celestial body gray image is used as the segmentation threshold T: Among them, g(u,v) is the grayscale value of the grayscale image, and H is the image height; The grayscale image is segmented by the segmentation threshold T to obtain the segmented image. The grayscale value s(u,v) of the segmented image is: Among them, the area with non-zero gray value is the target area, and the area with zero gray value is the background area.

4. A method for extracting the centroid of a small celestial body based on a joint pixel distribution frequency template as claimed in claim 3, characterized in that: The implementation method of step 2 is: Step 2.1: For the segmented image obtained in step 1, select the 8-connectivity criterion to mark the connected domains belonging to the same object in the segmented image, and based on the number of pixels in the connected domain, determine the threshold n. T Further distinguishing the background point area and one or more small celestial body illumination areas from the target area in step 1; The connected image with the label value l(u,v) is obtained through the 8-connectivity criterion. The area with the same label value is the connected domain belonging to the same object. It is known that the connected domain with the largest number of pixels belongs to the background area obtained in step 1, and the other connected domains belong to the target area; and the target area contains background point areas that may only occupy a few pixels, and one or more small celestial body illumination areas with a moderate number of pixels; Therefore, for different connected domains belonging to the target area, the number of pixels in each connected domain is n, and the judgment threshold n based on the number of pixels is set T , distinguish the background point area less than or equal to the judgment threshold and the small celestial body illumination area A greater than the judgment threshold from the target area; Step 2.2, combine all the small celestial body illumination areas obtained in step 2.1 to obtain a binary representation of the small celestial body combined connected image, further fill the local 0-value holes therein, and determine the final small celestial body connected image and the small celestial body connected domain C; By combining the illumination areas of all small celestial bodies, we can obtain a combined connected image of small celestial bodies represented by the binary value b(u,v): Further fill zero or more possible 0-value holes, that is, reassign all 0 values ​​in all holes to 1, and determine the final small celestial body connected image represented by the binary value c(u,v), and the small celestial body connected domain C corresponding to the value 1; Step 2.3: Use the small celestial body connected image in step 2.2 to perform logical judgment on the grayscale image in step 1, and obtain the grayscale value I used to reflect the brightness information of the small celestial body. c The connected grayscale image of the small celestial body (u, v) is used for the subsequent centroid extraction of the small celestial body.

5. The method for extracting the centroid of a small celestial body based on a joint pixel distribution frequency template as claimed in claim 4, characterized in that: The implementation method of step three is: Step 3.1: Based on the small celestial body connected domain obtained in step 2.2, the pixel distribution frequency templates of the small celestial body rows and columns in two directions are constructed, which are: Among them, N is the number of all valid pixels in the connected domain of the small celestial body, U and V are the sets of all pixels in the corresponding u-th row and v-th column respectively; Step 3.2: In both row and column directions, multiply and smooth the row and column distribution frequencies of corresponding pixels to obtain a joint pixel distribution frequency template for reflecting the shape information of small celestial bodies: The Hadamard product symbol ⊙ represents element-wise multiplication.

6. A method for extracting the centroid of a small celestial body based on a joint pixel distribution frequency template as claimed in claim 5, characterized in that: According to formula (6), the mass center coordinates of the small celestial body (u c ,v c )

Images