A method for extracting star points adapted to stray light interference

Abstract

The present invention discloses a star point extraction method adaptable to stray light interference. First, the first step is stray light suppression, where the background image containing stray light is processed to obtain a star map with a clean background. The second step is star map binarization, where it is determined whether the difference between the central pixel and the edge pixels meets the threshold requirements in both the horizontal and vertical directions, and binarization processing is performed. The third step is noise suppression, where the single pixel noise remaining in the star map is suppressed. The fourth step is star point marking, where the rough centroid coordinates of the star points under each connected domain are calculated. The fifth step is star point centroid calculation, where the accurate centroid coordinates of the star points in the 15#imgabs0#15 region are calculated using the centroid method with a threshold. The star point extraction method disclosed by the present invention can cope with the interference of stray light such as moonlight, sunlight, and earth atmosphere light, and can ensure the stable operation of the starlight navigation device during its on-orbit period.

Description

Technical Field

[0001] The present invention belongs to the field of target detection, and particularly relates to a star point extraction method adaptable to stray light interference. Background Art

[0002] As a high-precision and highly reliable on-orbit attitude measurement device, the starlight navigation device includes a relatively key technology, namely star point detection technology. The star point detection technology is to detect and calculate the centroid position of stars in the starry sky. However, in the actual use process, due to the existence of interference such as sunlight, moonlight, and earth atmosphere light (collectively referred to as stray light) in space, a large number of false targets will be extracted during the star point detection process, affecting subsequent recognition operations. Therefore, it is necessary to propose a reliable star point extraction method adaptable to stray light interference.

[0003] Currently, the star point detection methods for stray light interference include background prediction method, local adaptive threshold method, and connected component labeling method. However, these methods all have their own limitations, including high complexity, poor adaptability, or large consumption of storage space and operation time. Summary of the Invention

[0004] The technical problem to be solved by the present invention is: aiming at the deficiencies of the above traditional methods, the present invention proposes a star point detection method with low complexity, high real-time performance, and strong adaptability to stray light. This method uses background suppression, star Figure 2 thresholding, noise suppression, star point labeling, and centroid calculation to determine the star point positions in the star map in these 5 steps.

[0005] The technical solution adopted by the present invention is: a star point extraction method adaptable to stray light interference, specifically including the following steps:

[0006] Step 1: Input the star map Image with stray light interference;

[0007] Step 2: Stray light suppression. First, unify the highlighted pixels, set the pixels with gray values greater than the threshold Moon_Seg in the star map to the fixed value Moon_Seg, and obtain the image Image1;

[0008]

[0009] Considering that the star point targets are usually small in size, and the stray light presents the characteristics of large area and continuity, therefore, the operations of erosion first and then dilation under a large template are used to eliminate small targets such as star points first, and obtain the image Image2.

[0010]

[0011] Among them, SE1 is the erosion template; SE2 is the dilation template; is the erosion operation; For the dilation operation,

[0012] Through background subtraction, an image Image3 after stray light suppression can be obtained:

[0013]

[0014] Step 3: Star Figure 2 Thresholding. Starting from the horizontal and vertical dimensions, respectively, determine whether the sum of the central pixels and the sum of the edge pixels simultaneously satisfy the threshold relationship:

[0015]

[0016]

[0017] Where, , , are the 3 central pixels in the horizontal direction; , , are the 3 central pixels in the vertical direction; ~ are the pixels at the two side edges in the horizontal direction; ~ are the pixels at the two side edges in the vertical direction; is the threshold;

[0018] For the pixels that satisfy the threshold relationship, perform binary processing. After the entire star map is traversed, a binary star map Image4 can be obtained;

[0019] Step 4: Noise suppression. Adopt erosion and dilation operations under a small template to further suppress single-pixel noise.

[0020]

[0021] Step 5: Star point marking. Star point marking is performed using the 4-connected domain method based on the previous pixel for connected domain marking. The previous pixel represents the pixel in the previous column of the current pixel row and the pixel in the previous row. Use the relationship between the current pixel and the previous pixel to divide the connected domain. If the conditions are simultaneously met:

[0022]

[0023] Then it is marked as a new connected domain A(i,j) with a label number of m.

[0024] And for If one of the previous pixels is non-zero, then select the maximum value of the label m of the connected component C(i,j) among the previous pixels as the label of the connected component corresponding to the current pixel. And during the process of traversing the entire image, it is necessary to calculate the sum of the star point grayscales G(C) of the same connected component, the weighted value X(C) of the gray scale and the coordinate in the X direction, and the weighted value Y(C) of the gray scale and the coordinate in the Y direction:

[0025]

[0026]

[0027]

[0028] Among them, represents the connected component to which the pixel at the coordinate position (i,j) belongs; represents the connected component the sum of the accumulated grayscales, represents the gray scale value corresponding to the pixel at the coordinate position (i,j); represents the connected component the sum of the products of the gray scale and the horizontal coordinate i; represents the connected component the sum of the products of the gray scale and the vertical coordinate j,

[0029] When the entire image traversal is completed, the rough centroid position can be calculated:

[0030]

[0031]

[0032] Among them, represents the sum of the products of the gray scale and the horizontal coordinate corresponding to the nth connected component; represents the sum of the products of the gray scale and the vertical coordinate corresponding to the nth connected component; represents the sum of the accumulated grayscales corresponding to the nth connected component; and represents the rough centroid coordinates calculated for the nth connected component;

[0033] Step 6: Use the rough centroid coordinates calculated in Step 6 to calculate the precise coordinates in the original image Image1. Taking the rounded coordinates of the rough centroid as the origin, calculate the centroid coordinates of the star points within the 15 15 region.

[0034] The calculation method uses the centroid method with a threshold. First, calculate the average gray scale value of a total of 56 pixels in the edge region as the background value B, and then use 14 Subtract the background value B from each pixel value in the central region, and then multiply it by the coordinate value of the star point to obtain the accurate coordinates. :

[0035]

[0036]

[0037]

[0038] Further, the threshold Moon_Seg is usually set to 200.

[0039] Further, the erosion template corresponding to the large template erosion and dilation operation is 1 20, and the dilation template is 1 40. The erosion template usually needs to be larger than the size of the star point target.

[0040] Further, the line threshold Line_thred is usually set to 50.

[0041] Further, the erosion template corresponding to the small template erosion and dilation operation is 1 2, and the dilation template is 1 2.

[0042] Further, in step 2, the purpose of uniformizing the highlighted pixels is to make the uneven stray light background uniform, so as to only retain the position information of the stray light. This operation can ensure that after the background subtraction operation is completed, the stray light background is basically suppressed sufficiently.

[0043] Further, in step 4, the erosion and dilation operation of the small template is carried out. Due to the erosion operation carried out in step 2, the single-pixel noise (detector dead pixels or radiation noise) existing in the star map is covered together with the star points. Therefore, there may be a large number of separate bright pixels in the binary star map Image4. Therefore, the erosion and dilation operations under the small template are adopted to further suppress the single-pixel noise.

[0044] Compared with the prior art, the significant advantages of the present invention are:

[0045] (1) Strong adaptability to the stray light background. By uniformizing the background bright pixels and the background subtraction method, the influence of stray light can be effectively suppressed, and different stray light interferences can be adapted;

[0046] (2) High real-time performance. The execution process of the method mainly includes two erosion and dilation operations, background subtraction, and centroid calculation operations. The method has low complexity and is easy to implement on an embedded platform. Description of the Drawings

[0047] Figure 1 It is a schematic diagram of the work process;

[0048] Figure 2 It is a schematic diagram of horizontal and vertical threshold judgments;

[0049] Figure 3 It is a schematic diagram of 4-connected domains;

[0050] Figure 4 It is a schematic diagram of the centroid method with thresholds;

[0051] Figure 5 Star map Image with stray light interference;

[0052] Figure 6 Star map Image3 after stray light suppression;

[0053] Figure 7 Binarized star map Image4;

[0054] Figure 8 Star map Image5 after noise suppression;

[0055] Figure 9 Mark the detected star points in Image3. Specific implementation mode

[0056] The present invention will be further described below in conjunction with the accompanying drawings and specific implementation modes.

[0057] Combined with Figure 1 , the present invention is applicable to a star point extraction method adapted to stray light interference, and the specific steps are as follows:

[0058] Step 1: As Figure 5 shown, input the star map Image with stray light interference;

[0059] Step 2: Stray light suppression. First, unify the highlighted pixels, and set the pixels with gray values greater than the threshold Moon_Seg in the star map to the fixed value Moon_Seg to obtain the image Image1;

[0060]

[0061] Considering that the star point targets are usually small in size, and the stray light shows large area and continuity characteristics, the operations of erosion first and then dilation under a large template are used to eliminate small targets such as star points first to obtain the image Image2.

[0062]

[0063] Among them, SE1 is the erosion template; SE2 is the dilation template; is the erosion operation; This is the dilation operation,

[0064] As shown in Figure 6 , the image Image3 after stray light suppression can be obtained by background subtraction:

[0065]

[0066] Among them, pixels with negative values after subtraction are uniformly set to 0.

[0067] Step 3: Star Figure 2 Value quantization. As shown in Figure 2 , starting from the horizontal and vertical dimensions, respectively judge whether the sum of the central pixels and the sum of the edge pixels simultaneously satisfy the threshold relationship:

[0068]

[0069]

[0070] Among them, , , are the 3 central pixels in the horizontal direction; , , are the 3 central pixels in the vertical direction; ~ are the pixels at the two side edges in the horizontal direction; ~ are the pixels at the two side edges in the vertical direction; is the threshold,

[0071] For pixels that satisfy the threshold relationship, binary processing is performed. After the entire star map is traversed, as shown in Figure 7 , the binary star map Image4 can be obtained;

[0072] Step 4: Noise suppression. As shown in Figure 8 , the erosion and dilation operations under a small template are used to further suppress single-pixel noise.

[0073]

[0074] Step 5: Star point marking. As shown in Figure 3 , the star point marking is performed using the 4-connected domain method based on the previous pixels for connected domain marking. The previous pixels refer to the pixels in the previous column of the current pixel row and the pixels in the previous row. The connected domain is divided using the relationship between the current pixel and the previous pixels. If the following conditions are simultaneously met:

[0075]

[0076] It is then marked as the new connected component A(i, j) with the label number m.

[0077] For If there is a non-zero pixel among the previous pixels, then select the maximum value of the label numbers of the connected component C(i, j) in the previous pixels as the label number of the connected component corresponding to the current pixel. And during the process of traversing the entire image, it is necessary to calculate the sum of the star point grayscales G(C) of the same connected component, the weighted value of the gray scale and the coordinate in the X direction X(C), and the weighted value of the gray scale and the coordinate in the Y direction Y(C):

[0078]

[0079]

[0080]

[0081] Among them, represents the connected component to which the pixel at the coordinate position (i, j) belongs; represents the sum of the gray scale accumulations of the connected component represents the gray scale value corresponding to the pixel at the coordinate position (i, j); represents the sum of the accumulations of the products of the gray scale of the connected component represents the sum of the accumulations of the products of the gray scale of the connected component

[0082] When the entire image traversal is completed, the rough centroid position can be calculated:

[0083]

[0084]

[0085] Among them, represents the sum of the accumulations of the products of the gray scale and the horizontal coordinate corresponding to the nth connected component; represents the sum of the accumulations of the products of the gray scale and the vertical coordinate corresponding to the nth connected component; represents the sum of the gray scale accumulations corresponding to the nth connected component; and represents the rough centroid coordinates calculated for the nth connected component; Step 6: Using the rough centroid coordinates calculated in Step 6, calculate the exact coordinates in the original image Image1. Taking the rounded Coordinate of the rough centroid as the origin, calculate the centroid coordinates of the star points within the 15 × 15 area.

[0086] As Figure 4 shown, the calculation method uses the centroid method with a threshold.Figure 4 In this, M represents the central region image and N represents the edge region image. First, calculate the average gray value of 56 pixels in the edge region as the background value B, and then use 14 Subtract the background value B from each pixel value in the 14 central regions, and then multiply by the coordinate values of the star points to obtain the accurate coordinates :

[0087]

[0088]

[0089]

[0090] Figure 5 is the input star map with stray light interference, Figure 9 is the marking effect of the star points extracted by the method of the present invention on the star map after stray light suppression. It can be seen that the method of the present invention can effectively extract star point targets under the condition of stray light interference, including some bright stars covered by stray light. Moreover, no false targets are extracted in the stray light region by the method of the present invention. It can be proved that the star point extraction method described in the present invention has good extraction ability and robustness.

[0091] The parts not elaborated in the present invention belong to the well-known technologies in the art.

Claims

1. A star point extraction method adapted to stray light interference, characterized in that The method includes the following steps: Step 1: Input the star map Image with stray light interference; Step 2: Carry out stray light suppression; First, perform the unification of high-brightness pixels. Set the pixel points in the star map with gray values greater than the threshold Moon_Seg to the fixed value Moon_Seg to obtain the image Image1; ， Perform the operations of erosion first and then dilation under a large template to eliminate small star targets first, and obtain the image Image2, ， Among them, SE1 is the etching template; SE2 is the dilation template; is the etching operation; is the dilation operation, Through the method of background subtraction, obtain the image Image3 after stray light suppression, ； Step 3: Binarize the star map; Starting from the horizontal and vertical dimensions, respectively judge whether the sum of the central pixels and the sum of the edge pixels simultaneously satisfy the threshold relationship: ， ， Among them, , , is the center 3 pixels in the horizontal direction; , , is the center 3 pixels in the vertical direction; ~ are the pixels at the horizontal bilateral edges; ~ are the pixels at the vertical bilateral edges; is the threshold; Perform binarization processing on the pixels that satisfy the threshold relationship. After the entire star map is traversed, the binarized star map Image4 can be obtained; Step 4: Noise suppression; Use the operations of erosion and dilation under a small template to further suppress single-pixel noise, ， Among them, SE3 serves as both an etching template and a dilation template; is an etching operation; is a dilation operation; Step 5: Star point marking. The star point marking is carried out by using the four-connected domain method based on the previous pixel for connected domain marking. The previous pixel refers to the pixel in the previous column of the current pixel row and the pixel in the previous row. Use the relationship between the current pixel and the previous pixel to divide the connected domain. If the conditions are simultaneously met: ， Then mark it as a new connected domain A(i,j) and the mark number is m; Among them, represents the current pixel; represents the pixel in the column to the left of the current row; represents the pixel in the left column of the previous row; represents the pixel in the current column of the previous row; represents the pixel in the column to the right of the previous row; For If one of the previous pixels is not 0, then select the maximum value of the label m of the connected component C(i,j) among the previous pixels as the label of the connected component corresponding to the current pixel. During the process of traversing the entire image, it is necessary to calculate the sum of the star point grayscales G(C) of the same connected component, the weighted value X(C) of the gray scale and the coordinate in the X direction, and the weighted value Y(C) of the gray scale and the coordinate in the Y direction: ， ， ， Among them, represents the connected component to which the pixel at the coordinate position (i, j) belongs; represents the connected component of the sum of the gray levels accumulated; represents the gray level value corresponding to the pixel at the coordinate position (i, j); represents the connected component of the sum of the products of the gray level and the horizontal coordinate i accumulated; represents the connected component of the sum of the products of the gray level and the vertical coordinate j accumulated, When the entire image is traversed, calculate the rough centroid position: Among them, represents the cumulative sum of the product of the gray level corresponding to the nth connected component and the horizontal coordinate; represents the cumulative sum of the product of the gray level corresponding to the nth connected component and the vertical coordinate; represents the cumulative sum of the gray levels corresponding to the nth connected component; and represents the rough centroid coordinates calculated for the nth connected component; Step 6: Centroid calculation. Using the rough centroid coordinates calculated in Step 6, calculate the exact coordinates in the original image Image1. Taking the coordinates obtained by rounding the rough centroid as the origin, calculate the centroid coordinates of the star points within the 15 regions; The calculation method uses the centroid method with a threshold. First, calculate the average gray value of 56 pixels in the edge area as the background value B, and then use the 14 Subtract the background value B from each pixel value in the 14-center area, and then multiply it by the coordinate value of the star point to obtain the accurate coordinates , ， ， 。 2. The method for extracting star points adaptable to stray light interference according to claim 1, wherein: The threshold Moon_Seg described in Step 2 is usually set to 200.

3. A star point extraction method adapted to stray light interference according to claim 1, characterized in that: The etching template in Step 2 is 1 20, and the dilation template is 1 40 4. A star point extraction method for adapting to stray light interference according to claim 1, characterized in that: The line segment threshold Line_thred in Step 3 is usually set to 50.

5. A method for extracting star points adapted to stray light interference according to claim 1, characterized in that: The etching template in Step 4 is 1 2, and the dilation template is 1 2.

Images