Star point target detection method in complex background

Abstract

The present application relates to a kind of star point target detection method in complex background, comprising: a. the star point image of acquisition is carried out whole frame image scanning, whether the pixel point of the image is isolated noise point is judged, and isolated noise point is removed;B. the star point image of above-mentioned isolated noise point is removed is described using facet model, and the second derivative graph is calculated;C. the second derivative graph corresponding to the whole image is divided into regions, to confirm whether the region of division contains star point target;D. the pixel gray value in the region containing star point target is carried out gray value enhancement, and the pixel gray value in the region without star point target is carried out gray value weakening;E. after the image of step d processing, block and threshold segmentation are carried out, to detect star point target.The present application can detect star point target not submerged by stray light when complex background appears in star sensor field of view, and can stably output star sensor attitude information.

Description

Technical Field

[0001] This invention relates to the field of astronomical navigation technology, specifically to a method for detecting star targets against a complex background. Background Technology

[0002] A star sensor is a high-precision attitude-sensitive measurement instrument that determines the attitude of a carrier by detecting stars at different positions on the celestial sphere. It is currently the most widely used attitude measurement device in the aerospace field.

[0003] With the successful launch of numerous artificial satellites in the space environment, the imaging systems of star sensors are increasingly complex due to the interference of sunlight reflected by these satellites. Furthermore, there are issues such as interference from stray light from the moon, sun, and Earth's atmosphere.

[0004] Star sensors are usually equipped with sunshades to prevent stray light from the moon, sun, and earth's atmosphere from entering the detector target surface. However, in particularly harsh environments, stray light will always enter the sunshade and reach the detector target surface after multiple reflections. This results in a very complex background for the star point image after the star sensor is imaged. The star points are usually superimposed on a background with fluctuating gray levels and contain a lot of noise from isolated points such as thermal noise.

[0005] Under complex operating conditions, traditional star target detection methods cannot detect the position of stars, thus preventing star map recognition, attitude calculation, and other operations. This causes the star sensor to enter a "lost" state in space and lose its normal navigation function. Summary of the Invention

[0006] In view of this, it is necessary to provide a star target detection method under complex backgrounds, which can be applied to star sensors and solve the problem of star target detection failure under complex backgrounds caused by traditional star target detection methods.

[0007] This invention provides a method for detecting star targets in complex backgrounds. The method includes the following steps: a. Scanning the acquired star image frame by frame to determine whether the pixels in the image are isolated noise points and removing them; b. Describing the star image with isolated noise points removed using a facet model and calculating the second derivative map; c. Dividing the second derivative map of the entire image into regions to confirm whether the divided regions contain star targets; d. Enhancing the gray values ​​of pixels in regions containing star targets and weakening the gray values ​​of pixels in regions without star targets; e. Performing block segmentation and thresholding on the image processed in step d to detect star targets.

[0008] Preferably, step a includes:

[0009] The data processor uses a 3×3 template to scan the entire frame image and determines whether the gray value of the center pixel of the template is greater than the image threshold T1 and the gray values ​​of the surrounding 8 pixels are all less than the image threshold T1. If so, the center pixel is an isolated noise point and is replaced by the gray value of the right pixel or any of the surrounding pixels. If not, the center pixel is a non-noise point and its gray value remains unchanged.

[0010] Preferably, step b includes:

[0011] A facet model is used to describe a star-shaped image. The facet model uses a cubic equation in two variables to fit the grayscale value of the local neighborhood of the current pixel in the image.

[0012]

[0013] Where r and c are the row and column coordinates of a pixel in the local image, and k i represents the fitting coefficient.

[0014] Preferably, step b further includes:

[0015] Define R = {-2, -1, 0, 1, 2}, C = {-2, -1, 0, 1, 2}, (r,c)∈R×C. The second derivative is used to distinguish between star-shaped targets, stray backgrounds, and ordinary backgrounds. The second derivative of pixel f(r,c) with respect to the direction vector γ is expressed as:

[0016]

[0017] Where α represents the angle between the direction vector γ and the image row direction;

[0018] k i =f(r,c)*w i

[0019]

[0020]

[0021] Calculate the second derivative plots for directions 0°, 45°, 90°, and 135° using the second derivative.

[0022] Preferably, step c includes:

[0023] The second derivative map corresponding to the entire image is divided into regions, each region being 3×3 in size. A0, A1, and A2 are the mean values ​​of the second derivative within each region. The regional values ​​of the second derivative for different image features have the following characteristics: A0 is negative for star targets, A1 is positive, and A2 is positive; the mean values ​​of the second derivative in regions with ordinary background and stray light background are relatively close, which are completely different from the mean values ​​of the second derivative in regions with star targets.

[0024] Preferably, step d includes:

[0025] A new weighting coefficient Q is obtained using w4 and w6 to enhance the pixel grayscale values ​​in the region containing the target:

[0026] Q = -70(w⁴ + w⁶)

[0027] If a region contains a star-shaped target, satisfying the condition in step c, then the enhanced pixel grayscale value is the convolution of Q and the current pixel grayscale value:

[0028] if (A0 < 0 and A1 > 0 and A2 > 0)

[0029] When there are no star-shaped targets in a certain area, the grayscale value of this stray light background or ordinary background is reduced. It is half the grayscale value of the current pixel.

[0030] Preferably, step e includes:

[0031] Divide the image into sub-blocks with a resolution of 32×32 pixels. Then, calculate the image mean v within each sub-block, determine the values ​​of the row counter and column counter, and accumulate the grayscale sum for every 32 rows and 32 columns. Then, right-shift the grayscale sum by 10 bits at the falling edge of the field to obtain the mean of each sub-block. Calculate the image threshold T1 within each sub-block using the following formula:

[0032] T1=v+aσ

[0033] In the formula, v is the average gray value of the star map sub-block, a is the scale factor, and σ is the standard deviation of the star map sub-block.

[0034] Preferably, step e further includes:

[0035] Compare the size relationship between each pixel and the threshold. Pixels that are not smaller than the threshold are star targets, and pixels that are smaller than the threshold are background.

[0036] This application can identify and filter noise in star images against complex backgrounds, and is also applicable to star images against ordinary backgrounds. When a complex background appears in the star sensor's field of view, this application can detect star targets that are not obscured by stray light and can stably output the star sensor's attitude information. This invention is easy to implement on a hardware platform, and the method is simple, reliable, and effective, greatly improving the applicability of star sensors. Attached Figure Description

[0037] Figure 1 This is a flowchart of the star target detection method under complex backgrounds according to the present invention;

[0038] Figure 2 This is a schematic diagram of the second derivative plot region division in different directions according to the present invention. Detailed Implementation

[0039] The present invention will now be described in further detail with reference to the accompanying drawings and specific embodiments.

[0040] See Figure 1 The diagram shown is a flowchart of a preferred embodiment of the star target detection method under complex backgrounds of the present invention.

[0041] Step S1: Perform a full-frame image scan on the acquired star point image to determine whether any pixels in the image are isolated noise points, and remove isolated noise points. Specifically:

[0042] The star image captured by the star sensor typically contains isolated noise points. The data processor scans the entire frame using a 3×3 template, determining whether the grayscale value of the center pixel is greater than the image threshold T1 (the method for determining the value of T1 is described in step S5), and whether the grayscale values ​​of the surrounding 8 pixels are all less than the image threshold T1. If so, the center pixel is considered an isolated noise point, and its grayscale value is replaced by the grayscale value of the pixel to its right or any of the surrounding pixels. If not, the center pixel is considered a non-noise point, and its grayscale value remains unchanged. This process is repeated across the entire frame to remove isolated noise points.

[0043] The important role of this step is that a large number of isolated noise points will occupy the computing and storage resources of the data processor. Removing isolated noise points from the star image first can greatly reduce the amount of computation in data processing, prevent data blockage, and improve the computing efficiency of the data processor.

[0044] Step S2 involves using a facet model to describe the star image after removing isolated noise points and calculating the second derivative plot. Specifically:

[0045] The facet model is used to describe the star image. The facet model uses a cubic equation to fit the gray level in the local neighborhood of the current pixel in the image.

[0046]

[0047] Where r and c are the row and column coordinates of a pixel in the local image, and k i The fitting coefficients are denoted as .

[0048] Define R = {-2, -1, 0, 1, 2}, C = {-2, -1, 0, 1, 2}, and (r, c) ∈ R × C. This application utilizes the second derivative value to distinguish the features of star-shaped targets, stray light backgrounds, and ordinary backgrounds. The second derivative of pixel f(r, c) with respect to the direction vector γ is expressed as:

[0049]

[0050] Where α represents the angle between the direction vector γ and the image row direction.

[0051] k i =f(r,c)*w i

[0052]

[0053]

[0054] Calculate the second derivative plots for directions 0°, 45°, 90°, and 135° using the second derivative.

[0055] Step S3 involves dividing the second derivative graph of the entire image into regions to determine whether each region contains star-shaped targets. Specifically:

[0056] Divide the second derivative plot of the entire image into regions, each region being 3×3 in size. Please refer to the region division rules for different directions. Figure 2 A0, A1, and A2 represent the mean of the second derivative within each region. The regional values ​​of the second derivative for different image features exhibit the following characteristics: A0 is negative for star-shaped targets, A1 is positive, and A2 is also positive. In contrast, the mean values ​​of the second derivative in regions with ordinary backgrounds and stray light backgrounds are quite similar, completely different from the mean values ​​of the second derivative in regions with star-shaped targets. Therefore, this characteristic can be used to confirm whether a region contains star-shaped targets.

[0057] Step S4: Enhance the grayscale values ​​of pixels in areas containing star-shaped targets, and reduce the grayscale values ​​of pixels in areas without star-shaped targets. Specifically:

[0058] The grayscale values ​​of pixels in areas with star-like features are enhanced, while the grayscale values ​​of pixels in areas without star-like features are weakened. Specifically, a new weighting coefficient Q is obtained using w4 and w6 to enhance the grayscale values ​​of pixels within the target area.

[0059] Q = -70(w⁴ + w⁶)

[0060] If a region contains a star-shaped target, satisfying the condition of containing a target in step S3, then the enhanced pixel grayscale value is the convolution of Q and the current pixel grayscale value:

[0061] if (A0 < 0 and A1 > 0 and A2 > 0)

[0062] When there are no star-shaped targets in a certain area, the grayscale value of this stray light background or ordinary background is reduced. It is half the grayscale value of the current pixel.

[0063] Step S5 involves segmenting the image processed in step S4 into blocks and applying thresholds to detect star-shaped targets. Specifically:

[0064] The processed image obtained in step S4 By performing block segmentation and thresholding, the target can be detected. The specific implementation is as follows: Each sub-block is divided into 32×32 pixels. Then, the image mean *v* within each sub-block is calculated. The values ​​of the row and column counters are checked, and the grayscale sum is accumulated for every 32 rows and 32 columns. Then, at the falling edge of the field, the grayscale sum is right-shifted by 10 bits to obtain the mean of each sub-block. The image threshold *T1* within each sub-block is calculated using the following formula:

[0065] T1=v+aσ

[0066] In the formula, v is the average gray value of the star map sub-block, a is the scale factor, and σ is the standard deviation of the star map sub-block. Based on the properties of the normal distribution and referring to experimental results, a is taken as 3.

[0067] Finally, the size relationship between each pixel and the threshold is compared. Pixels that are not smaller than the threshold are star targets, and pixels that are smaller than the threshold are background.

[0068] Although the present invention has been described with reference to the present preferred embodiments, those skilled in the art should understand that the above preferred embodiments are only used to illustrate the present invention and are not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for detecting star-shaped targets under complex backgrounds, characterized in that, The method includes the following steps: a. Perform a full-frame image scan on the acquired star point image, determine whether the pixels in the image are isolated noise points, and remove isolated noise points; b. Use a facet model to describe the star image above after removing isolated noise points, and calculate the second derivative plot; c. Divide the second derivative plot of the entire image into regions to confirm whether the divided regions contain star-shaped targets; d. Enhance the grayscale values ​​of pixels in areas containing star targets, and reduce the grayscale values ​​of pixels in areas without star targets; e. The image processed in step d is segmented into blocks and thresholded to detect star targets.

2. The method as described in claim 1, characterized in that, Step a includes: The data processor uses a 3×3 template to scan the entire frame image and determines whether the gray value of the center pixel of the template is greater than the image threshold T1 and the gray values ​​of the surrounding 8 pixels are all less than the image threshold T1. If so, the center pixel is an isolated noise point and is replaced by the gray value of the right pixel or any of the surrounding pixels. If not, the center pixel is a non-noise point and its gray value remains unchanged.

3. The method as described in claim 2, characterized in that, Step b includes: A facet model is used to describe a star-shaped image. The facet model uses a bivariate cubic equation to fit the grayscale of the local neighborhood of the current pixel in the image. Where r and c are the row and column coordinates of a pixel in the local image, and k i The fitting coefficients are denoted as .

4. The method as described in claim 3, characterized in that, Step b further includes: Define R = {-2, -1, 0, 1, 2}, C = {-2, -1, 0, 1, 2}, (r,c)∈R×C. The second derivative is used to distinguish between star-shaped targets, stray backgrounds, and ordinary backgrounds. The second derivative of pixel f(r,c) with respect to the direction vector γ is expressed as: Where α represents the angle between the direction vector γ and the image row direction; k i =f(r,c)*w i Calculate the second derivative plots for directions 0°, 45°, 90°, and 135° using the second derivative.

5. The method as described in claim 4, characterized in that, Step c includes: The second derivative map corresponding to the entire image is divided into regions, each region being 3×3 in size. A0, A1, and A2 are the mean values ​​of the second derivative within each region. The regional values ​​of the second derivative for different image features have the following characteristics: A0 is negative for star targets, A1 is positive, and A2 is positive; the mean values ​​of the second derivative in regions with ordinary background and stray light background are relatively close, which are completely different from the mean values ​​of the second derivative in regions with star targets.

6. The method as described in claim 5, characterized in that, Step d includes: A new weighting coefficient Q is obtained using w4 and w6 to enhance the pixel grayscale values ​​in the region containing the target: Q = -70(w⁴ + w⁶) If a region contains a star-shaped target, satisfying the condition in step c, then the enhanced pixel grayscale value is the convolution of Q and the current pixel grayscale value: if(A0<0and A1>0and A2>0) When there are no star-shaped targets in a certain area, the grayscale value of this stray light background or ordinary background is reduced. It is half the grayscale value of the current pixel.

7. The method as described in claim 6, characterized in that, Step e includes: Divide the image into sub-blocks with a resolution of 32×32 pixels. Then, calculate the image mean v within each sub-block, determine the values ​​of the row counter and column counter, and accumulate the grayscale sum for every 32 rows and 32 columns. Then, right-shift the grayscale sum by 10 bits at the falling edge of the field to obtain the mean of each sub-block. Calculate the image threshold T1 within each sub-block using the following formula: T1=v+aσ In the formula, v is the average gray value of the star map sub-block, a is the scale factor, and σ is the standard deviation of the star map sub-block.

8. The method as described in claim 7, characterized in that, Step e further includes: Compare the size relationship between each pixel and the threshold. Pixels that are not smaller than the threshold are star targets, and pixels that are smaller than the threshold are background.

Images