High-real-time daytime stray light star map star point extraction method

Abstract

The invention relates to a high-real-time daytime stray light star map star point extraction method, which comprises the following steps of: firstly, carrying out stray light background suppression on a star map by adopting an improved soft morphological filtering algorithm; and removing residual single-point noise from the image subjected to stray light suppression by adopting a morphological opening operation. Secondly, bright pixels are extracted from the filtered star map, four-connected-domain judgment is carried out through a clustering method, the connected domains are ranked from high to low according to the total energy, and four stars with the highest energy are selected to carry out triangle recognition matching based on a bright star table; thirdly, postures are calculated based on the successfully matched star points, and the star points in the view field are back-projected to an original star image surface in combination with a complete star catalogue; and finally, windowing an original star map according to the calculated star point image plane coordinates, performing stray light suppression on a window image by adopting an edge pixel fitting method, and completing star point extraction by adopting adaptive threshold segmentation and a four-connected-domain judgment algorithm.

Description

Technical Field

[0001] The invention relates to a high-real-time daylight stray light star map star point extraction method, which is applied to star map processing of near-ground platforms such as airplanes, unmanned aerial vehicles, and ships. Background Art

[0002] The all-day star sensor is used in a strong stray light background. The imaging environment is complex and changeable, and it is easily interfered by background stray light and pseudo stars from floating objects in the air. On the one hand, a large number of floating objects in the air enter the field of view during flight, and the bright spots formed by the reflected sunlight will bring a complex stray light environment. At the same time, the size and shape of the accompanying floating objects are similar to star points, forming a large number of pseudo star points. On the other hand, the background brightness in the atmosphere is higher than that in the space environment, and the contrast of the star point target is low, which is relatively more difficult to extract.

[0003] The star point extraction of the daytime stray light star map of the all-day star sensor mainly needs to solve the problems of stray light image noise and background suppression, and the identification of pseudo-star points under daytime conditions. At present, the research on star point extraction under daytime stray light environment at home and abroad mainly focuses on the background suppression of stray light star maps of star sensors on medium and low altitude platforms such as airborne platforms, and there is little research on methods for pseudo-star processing.

[0004] Traditional stray light image background suppression methods mainly include multi-frame accumulation and averaging, median filtering, high-pass filtering, morphological filtering, etc. The multi-frame accumulation and averaging method adds the grayscale of the corresponding pixels of multiple frames of star images taken continuously and then calculates the average value. This method can eliminate random noise, but it has little effect under the strong shot noise caused by daytime stray light; median filtering replaces the value of the template center point with the median of each point in the template, which can effectively suppress impulse noise, but the effect on normal white noise is not obvious. At the same time, median filtering will produce pixel offsets, affecting the accuracy of centroid positioning; high-pass filtering performs Fourier transform on the star map, retaining the frequency components above the cutoff frequency in the star map, which can be used to eliminate low-frequency noise, but the noise suppression effect on stray light background images is not obvious, and the processing process is computationally intensive and the real-time performance is poor; morphological filtering is based on structural elements, measures and extracts corresponding shapes in the image, and can effectively suppress complex star map backgrounds, but there are still a lot of clutter residues in the filtering results, which need further processing.

[0005] Under the complex stray light background of the day, floating objects in the sky and residual pre-processing filters will cause a large number of pseudo star interference in the star map. The traditional method is to extract all of them first and then match and identify them. This method requires a lot of storage space and more than ten seconds of computing time, which has a serious impact on the real-time extraction. In order to identify pseudo stars under complex backgrounds, the pseudo star identification method of multi-frame comparison is currently mainly used. Star matching is performed based on whether the pixel position changes of two connected domains in the star map of two frames with different shooting times conform to the law of motion changes of star points. This method can effectively avoid the interference of pseudo star points caused by floating objects in the sky, but it is difficult to achieve for scenes where the positions between frames remain almost unchanged, and the multi-frame comparison calculation amount is large in the case of a large number of pseudo star points, which affects the real-time performance and accuracy. Summary of the invention

[0006] The technical problem to be solved by the present invention is to overcome the deficiencies of the prior art and achieve high-precision and real-time recognition and extraction of star points.

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

[0008] In a first aspect, the present invention provides a high-real-time daytime stray light star map star point extraction method, comprising:

[0009] The improved soft morphological filtering algorithm is used to suppress the stray light background of the star image. The morphological opening operation is used to remove the residual single point noise of the image with suppressed stray light background to obtain the filtered star image.

[0010] Extract bright pixels from the filtered star map and use clustering method to judge the four-connected domain. Sort the connected domains from high to low according to the total energy, and select the four stars with the highest energy for triangle recognition and matching based on the bright star list.

[0011] Calculate the pose based on the successfully matched star points, and back-project the star points in the field of view onto the original star image plane in combination with the complete star catalog;

[0012] The original star map is windowed according to the calculated image coordinates of the star points, and the stray light of the window image is suppressed by the edge pixel fitting method. The star points are extracted using the adaptive threshold segmentation and four-connected domain judgment algorithm.

[0013] In a second aspect, the present invention provides a computer-readable storage medium having computer program instructions stored thereon. When the computer program instructions are loaded and executed by a processor, the processor executes the high-real-time daylight stray light star map star point extraction method described in the first aspect.

[0014] Compared with the prior art, the present invention has the following beneficial effects:

[0015] (1) The present invention proposes a method of using a bright star list and a partially connected domain sequence with the highest grayscale value for rapid matching and identification, and then back-projecting to extract the remaining dim star points, thereby avoiding the extraction and identification process of hundreds of pseudo stars, greatly reducing the amount of calculation, shortening the working time and storage space of star point identification and matching, achieving high-real-time star point identification and matching, and effectively solving the problem of star point extraction in an environment where a large number of pseudo star points are caused by accompanying floating objects.

[0016] (2) The present invention aims at suppressing the complex star map background formed by sunlight reflection in the daytime environment and the single point noise remaining after filtering. The improved soft morphological filtering and morphological opening operation method are adopted to effectively remove the influence of stray light background and suppress single point noise, highlight the contrast between the target area and the surrounding area, and significantly improve the probability of star point target extraction.

[0017] (3) The present invention adopts a method of window background fitting to eliminate background stray light, thereby solving the problem of accuracy of star point extraction in stray light environments and achieving high-precision star point extraction in complex light background environments.

[0018] (4) The present invention can be extended to the extraction of faint star points in various environments. For sensors outside the atmosphere, this method can be used to detect star points with weaker energy. The bright star list is used to quickly identify and match, and then the dark star list is back-projected and the window image is intercepted. The background fitting is performed to finally extract the star points, which can quickly and accurately realize the detection of faint star points in space.

[0019] (5) The present invention is easy to implement in hardware, has a small amount of calculation, good real-time performance, and has high use value, and can be widely used in all-day star sensors. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] Figure 1 This is a flow chart of the algorithm for extracting star points from high-real-time daytime stray light star maps;

[0021] Figure 2(a) is an improved soft morphological dilation filter structure element b;

[0022] Figure 2(b) shows the improved soft morphological corrosion filter center structure element B1;

[0023] Figure 2(c) shows the improved soft morphological corrosion filter edge structure element B2;

[0024] Figure 3 It is the structural element of the morphological opening sequence;

[0025] Figure 4 This is the back-projection result diagram of star points within the detectable magnitude range. DETAILED DESCRIPTION

[0026] To make the objectives, technical solutions and advantages of the present invention more clear, the embodiments of the present invention will be further described in detail below with reference to the accompanying drawings.

[0027] like Figure 1 As shown, a high-real-time daytime stray light star map star point extraction method comprises:

[0028] (1) An improved soft morphological filtering method is used to suppress stray light background in the star image and highlight the pixel grayscale difference between the target area and the surrounding area. The specific star image processing process is as follows:

[0029] (1.1) Select the edge filter template to perform soft morphological dilation on the original star image row by row and column by column to obtain the first step filter value at the center of the template. Select the size of the structure element to be p×p, the center (p-4)×(p-4) range of the structure element to be 0, and the pixel weight of the edge ring range to be 1. In the soft morphological filtering, the structure element is divided into the central structure element b1 and the edge structure element b2. Therefore, during filtering, the central structure element is the center (p-4)×(p-4) range of the original structure element, and the edge structure element is the center (p-4)×(p-4) blank p×p ring range.

[0030] In this embodiment, a 11×11 structuring element b is selected, as shown in FIG2(a), the center 7×7 is blank, and the pixel weight of the edge ring range is 1. In soft morphological filtering, the center structuring element b1 is considered to be a 7×7 structuring element at the center of the structuring element b, and the edge structuring element b2 is considered to be a 11×11 ring structuring element with a center 7×7 blank. The grayscale value at the center of the template is replaced by the result of the expansion processing at other positions, that is, the maximum value within the ring range of the structuring element b2. The original stray light star image is subjected to the soft morphological expansion operation to obtain the first step filtered image.

[0031] (1.2) Select the filter template and perform soft morphological corrosion on the first step filtered image row by row and column by column to obtain the stray light background value at the center of the template. The size of the structure element is the same as that in step (1.1), which is p×p. The pixel weight of the (p-4)×(p-4) range in the center of the structure element is k (1≤k≤min{p×p / 2,(p-4)×(p-4)}), and the pixel weight of the edge ring range is 1. Use the above structure element to perform corrosion operation on the first step filtered image. The grayscale of the template center is represented by the kth value of the structure element grayscale in ascending order to obtain the stray light background grayscale of the stray light star map.

[0032] The weight of the central structural element is selected according to the grayscale of the image. The grayscale mean f1 of the central structural element and the grayscale mean f2 of the edge structural element are calculated respectively. The difference between f1 and f2 is calculated and compared with min{f1,f2}. When the difference is less than t times min{f1,f2}, the k value is 1; when the difference is greater than t times min{f1,f2}, the k value is calculated according to the current pixel grayscale. The t value is a threshold preset according to the uniformity of the ambient stray light background, 0<t≤1. The greater the background fluctuation, the higher the threshold is set, which can generally be set to 0.1.

[0033]

[0034] In the formula, k(i,j) represents the weight value at the (i,j) pixel, and f(i,j) represents the image grayscale value at the (i,j) pixel. The structural elements B are sorted in ascending order according to the weight of the central structural element pixel, and the kth value is selected as the grayscale value of the structural element center after the corrosion operation. When |f1-f2|>tmin{f1,f2}, k(i,j) is averaged to obtain the k value.

[0035] In this embodiment, a 11×11 structuring element B is selected, the central structuring element B1 is a 7×7 range at the center of the original structuring element B, and the pixel weight is k (1≤k≤49), as shown in Figure 2(b); the edge structuring element B2 is a 11×11 ring structuring element with a 7×7 blank in the center, and the pixel weight is 1, as shown in Figure 2(c). The above operation is performed row by row and column by column on the first step filtered image to obtain the second step filtered image.

[0036] (1.3) Subtract the grayscale value of the original stray light star image from the grayscale value of the second-step filtered image, and calculate the difference at the same pixel position. If the grayscale value of the original star image is less than the filtered value, the grayscale is set to 0 to obtain the star point image after suppressing the stray light background.

[0037] (2) Perform morphological opening operation on the filtered star image to remove single point noise in the image and smooth the target edge.

[0038] Select a 2×1 column structure element to perform morphological opening operation on the filtered image. The structure element is Figure 3 As shown, the abnormal pixels in the image are eliminated. First, the column structure element is moved on the filtered image, and the center value of the structure element is the minimum grayscale value of all pixels in the template to obtain the minimum value image. Secondly, the column structure element is moved on the minimum value image, and the center value of the structure element is the maximum grayscale value of all pixels in the template to obtain the filtered star map with noise removed.

[0039] (3) Extract, identify, and match the real star points in the image in combination with the star sensor star catalog, and back-project the star points in the star catalog onto the image plane according to the current attitude.

[0040] Traverse the processed star map and extract the M pixels with the highest grayscale values ​​in the image (M is preset according to the actual flight environment, generally 500≤M≤5000). Use the clustering method to determine the four-connected domains of the extracted bright pixels, and determine the connected domains with a total area greater than 4. Considering the imaging of the star points on the image plane, eliminate the connected domains that do not conform to the shape and size of the star points, and sort the remaining connected domains according to the total energy grayscale value.

[0041] Select the 4 connected domains with the highest grayscale values ​​and perform triangle recognition in combination with the star sensor's bright star catalog (generally a catalog of stars with magnitudes below 4.5); if the recognition is successful, stop matching; if it is unsuccessful, add the 5th connected domain for recognition, and so on, until 4 stars are successfully recognized.

[0042] Based on the 4 identified star points, the attitude is roughly calculated. According to the current attitude, the star points in the complete star catalog (based on the star sensor detection sensitivity, such as the 7th magnitude star catalog) within the field of view of the star sensor are back-projected into the image to obtain the coordinates of the star points on the image plane. The back-projection results of the star points within the detectable magnitude range are as follows: Figure 4 As shown, the positions of the stars in the image and their instrumental magnitudes are marked in the figure.

[0043] (4) According to the image plane coordinates of the star points, the original star map is windowed, and the stray light of the window image is suppressed by edge pixel fitting to achieve star point extraction and attitude calculation. The specific steps are as follows:

[0044] (4.1) According to the image plane coordinates of the star points obtained in step (3), a window is opened in the original star map to obtain a window image of size Npix×Npix (usually set to 24×24).

[0045] (4.2) Fit the edge pixels of each window image to obtain the stray light background, and subtract them to achieve the stray light suppression of the star point window image. The window image background fitting can adopt the edge pixel mean fitting method, information entropy fitting method, etc. In the present invention, the least squares method is used to fit the window edge pixels.

[0046] Specifically, the window image stray light fitting method based on the least squares method is:

[0047] Assume that the edge of the window image consists of n pixels, where the i-th edge pixel is at the y-th i Row x i Column, whose pixel gray value is E i The fitted stray light background surface is where a i is the fitting coefficient of the stray light background surface, let Then X=[a1 a2 a3 a4 a5 a6] T is the fitting result, so we have

[0048]

[0049] From X = A -1 b can be solved to get the fitting coefficient of the stray light background, and the stray light background map of the entire window can be obtained according to the fitting result. The grayscale of the window image is subtracted from the stray light background map at the same position. If the window grayscale is less than the stray light background grayscale, it is 0, and the window star map without the stray light background is obtained.

[0050] (4.3) For each window image after removing the stray light background, adaptive threshold segmentation is used to select the target star point area, and the star point distribution is determined based on the four-connected domain, and the star point centroid position is calculated to achieve high-precision star point recognition and extraction under the stray light background.

[0051] The contents not described in detail in the specification of the present invention belong to the common knowledge of those skilled in the art.

[0052] Although the present invention has been disclosed as above in the form of a preferred embodiment, it is not intended to limit the present invention. Any person skilled in the art may make possible changes and modifications to the technical solution of the present invention by using the methods and technical contents disclosed above without departing from the spirit and scope of the present invention. Therefore, any simple modifications, equivalent changes and modifications made to the above embodiments based on the technical essence of the present invention without departing from the content of the technical solution of the present invention shall fall within the protection scope of the technical solution of the present invention.

Claims

1. A high-real-time method for extracting star points from daylight stray light star maps. It is characterized in that include: An improved soft morphological filtering algorithm is used to suppress stray light background in star images. The morphological opening operation is used to remove the residual single-point noise from the image with suppressed stray light background to obtain the filtered star map; Extract bright pixels from the filtered star map and use clustering method to judge the four-connected domain. Sort the connected domains from high to low according to the total energy, and select the four stars with the highest energy for triangle recognition and matching based on the bright star list. Calculate the pose based on the successfully matched star points, and back-project the star points in the field of view onto the original star image plane in combination with the complete star catalog; The original star map is windowed according to the calculated image coordinates of the star points, and the stray light of the window image is suppressed by the edge pixel fitting method. The star points are extracted using the adaptive threshold segmentation and four-connected domain judgment algorithm.

2. The high-real-time daylight stray light star map star point extraction method according to claim 1, It is characterized in that The improved soft morphological filtering algorithm is used to suppress stray light background of star images, including: Select the edge filter template to perform soft morphological dilation on the original star image row by row and column by column to obtain the first step filtered image at the center of the template; Select a filter template to perform soft morphological corrosion on the first step filtered image row by row and column by column to obtain the stray light background value at the center of the template; The original star image and the stray light background value are used to obtain the star point image after suppressing the stray light background.

3. The high-real-time daylight stray light star map star point extraction method according to claim 2, It is characterized in that Soft morphology expansion processing includes: The size of the structural element is selected as p×p, the range of the structural element center (p-4)×(p-4) is 0, and the pixel weight of the edge ring range is 1; in the soft morphological filtering, the structural element is divided into the central structural element b 1 and edge structure element b 2 , the central structural element is the center (p-4)×(p-4) range of the original structural element during filtering, the edge structural element is the center (p-4)×(p-4) blank p×p ring range, and the gray value of the center position of the edge filter template is replaced by the result of the expansion processing of other positions, that is, the structural element b 2 The maximum value within the annular range.

4. The high-real-time daylight stray light star map star point extraction method according to claim 3, It is characterized in that Soft morphological etching treatments include: The pixel weight of the (p-4)×(p-4) range at the center of the structure element is k (1≤k≤min{p×p / 2,(p-4)×(p-4)}), and the pixel weight of the edge ring range is 1; the first step filtered image is eroded using the above structure element, and the grayscale of the center position of the edge filter template is represented by the kth value of the structure element grayscale in ascending order to obtain the stray light background value of the stray light star map; The weight of the structure element center is selected according to the image grayscale, and the grayscale mean f of the structure element center is calculated respectively. 1 and the grayscale mean f of the edge structure element 2 , calculate f 1 With f 2 The difference between 1 ,f 2 }, when the difference is less than t times min{f 1 ,f 2 }, the k value is 1; when the difference is greater than t times the min{f 1 ,f 2 }, the k value is calculated based on the current pixel grayscale, and the t value is a threshold preset based on the uniformity of the ambient stray light background; In the formula, k(i,j) represents the weight value at the (i,j) pixel, and f(i,j) represents the image grayscale value at the (i,j) pixel. The structural elements B are sorted in ascending order according to the weight of the central structural element pixel, and the kth value is selected as the grayscale value of the structural element center after the corrosion operation. When |f 1 -f 2 |>tmin{f 1 ,f 2 }, take the average of k(i,j) to get the k value.

5. The high-real-time daylight stray light star map star point extraction method according to claim 2, It is characterized in that Subtract the grayscale of the original star image from the stray light background value, and calculate the difference at the same pixel position. If the grayscale value of the original star image is less than the stray light background value, the grayscale is set to 0 to obtain the star point image after suppressing the stray light background.

6. The high-real-time daylight stray light star map star point extraction method according to claim 1, It is characterized in that The morphological opening operation is used to remove the residual single point noise of the image with suppressed stray light background, including: A 2×1 column structure element is selected to perform morphological opening operation on the image with suppressed stray light background. First, the column structure element is moved on the image with suppressed stray light background, and the center value of the column structure element is the minimum grayscale value of all pixels in the template to obtain the minimum value image; then the column structure element is moved on the minimum value image, and the center value of the column structure element is the maximum grayscale value of all pixels in the template to obtain the filtered star map with noise removed.

7. The high-real-time daylight stray light star map star point extraction method according to claim 1, It is characterized in that When the clustering method is used to judge the four-connected domain, all connected domains with an area greater than 4 are determined, and the connected domains that do not conform to the shape and size of the star point are eliminated.

8. The high-real-time daylight stray light star map star point extraction method according to claim 1, It is characterized in that Window image background fitting methods include edge pixel mean fitting method, information entropy fitting method or least squares method.

9. A computer-readable storage medium having computer program instructions stored thereon, wherein when the computer program instructions are loaded and executed by a processor, the processor is caused to execute the method according to any one of claims 1 to 8.

Images