An adaptive distance-power exponential iterative weighted centroid method for corner reflector image point fitting

Abstract

The application discloses a corner reflector image point fitting adaptive distance-power exponential iterative weighted centroid method. On the basis of solving the position of the corner reflector image point by the distance Doppler model method, the coarse position of the corner reflector image point on the SAR image is determined by artificial visual interpretation and analysis of the surrounding ground feature appearance; the window size of the corner reflector image point on the SAR image is judged according to the pixel value, and the fitting window is adaptively judged based on the window size; the pixel value and the coordinates of all the image elements in the fitting window are searched and recorded; each image element searched is subdivided, and the pixel value and the coordinates are recorded; the distance-power exponential iterative weighted centroid model is constructed according to the distance relationship between the pixels to accurately extract the corner reflector point; finally, the precision evaluation method of the distance-power exponential iterative weighted centroid model for extracting the corner reflector image point is provided.

Description

TECHNICAL FIELD

[0001] The application belongs to the field of satellite calibration, and particularly relates to a precision extraction analysis method for corner reflector points on SAR images. BACKGROUND

[0002] The extraction precision of corner reflector image points directly affects the geometric calibration effect of a synthetic aperture radar (SAR). The traditional centroid method for extracting corner reflector image points on SAR images has defects in positioning precision and speed. The distance-square-weighted centroid method for extracting corner reflector image points takes the square of distance as a weighting function, so that the pixels closer to the distance spot center have greater weights, and the pixels farther from the distance spot center have less influence on the centroid judgment, thereby weakening the noise far from the spot center to reduce the interference of noise on the detection of corner reflector points and improve the detection precision of corner reflector points.

[0003] The position of a corner reflector point on a SAR image is usually acquired through manual visual interpretation and collection. The original SAR image is limited by resolution or unclear details, so that the precision of the extracted corner reflector image point position is limited. At present, more design experiments of corner reflectors are carried out, but the research on the extraction of corner reflector points on SAR images is less. The extraction of corner reflector points on SAR images has been researched at home and abroad. Dr. Shaye of Germany and Dr. Xue Xiaorong of Peking University have done research on the identification of corner reflectors on SAR images. Their work is only of reference for themselves and is not universal for all corner reflector points. Song Ruiqing et al. use the square-weighted centroid method to realize fine positioning of point targets in SAR images, but the precision of the method can only reach the centimeter level.

[0004] In summary, in view of the demand for the extraction precision of corner reflector image points on SAR images, a distance-square-weighted centroid method is proposed by constructing a distance-square-weighted formula according to the distance relationship between corner reflector image points through a fitting window of a certain size. SUMMARY

[0005] The application proposes a distance-square-weighted centroid method for fitting corner reflector image points in view of the demand for the extraction precision of corner reflector points on SAR images. The application can be applied to the field of precision extraction of corner reflector points on SAR images.

[0006] To achieve the above purpose, the application comprises the following steps:

[0007] S1: determining the coarse position of the corner reflector point on the SAR image. A certain size window is used to search for the maximum pixel value of the "bright spot" near the corner reflector image point on the SAR image calculated by the range Doppler model method, and then the surrounding ground features are analyzed and compared by artificial visual interpretation to determine the maximum pixel value position as the coarse position of the corner reflector image point;

[0008] S2: determining the window pixel information near the coarse position. A certain size pixel window is selected as the fitting window with the coarse position determined in S1 as the center, and the pixel values and corresponding coordinates of all pixels in the window are extracted in order;

[0009] S3: constructing the distance square weighted centroid method based on the distance relationship between pixels to accurately extract the corner reflector point;

[0010] S4: accuracy evaluation. The position of the corner reflector point calculated in S3 is compared with the position of the corner reflector image point calculated by the range Doppler model method, and the accuracy error of the corner reflector point extracted by the adaptive weighted centroid method is analyzed and compared.

[0011] Further, step S1 includes the following steps:

[0012] (1) A certain size window is used to search for the maximum pixel value of the "bright spot" near the corner reflector image point on the SAR image calculated by the range Doppler model method;

[0013] (2) Then the surrounding ground features are analyzed and compared by artificial visual interpretation to determine the maximum pixel value position as the coarse position of the corner reflector image point;

[0014] Further, step S2 includes the following steps:

[0015] (1) A certain size pixel window is selected as the finding window of the high-precision corner reflector point coordinates with the coarse position determined in step S1 as the coordinates;

[0016] (2) The pixel values and coordinates of each pixel obtained according to the finding window are recorded in order;

[0017] Further, step S3 includes the following steps:

[0018] (1) According to the distance weighted centroid method formula, the reciprocal of the distance is taken as the weighting function to calculate the position of the corner reflector image point. The formula is as follows:

[0019]

[0020] In the formula, (x, y) is the coordinates of the current measured pixel, (x0, y0) is the center coordinates of the corner reflector "bright spot", The center coordinates of the calculated corner reflector "bright spot" are I(x, y), and the pixel value of the current pixel is I(x, y).

[0021] (2) Since the gray value distribution of the corner reflector image point is Gaussian distribution, the gray value closer to the center is larger, so the distance weighted centroid method uses the square of the distance to increase the contribution of the pixels close to the center to the calculation of the centroid. The formula is as follows:

[0022]

[0023] In the formula, (x, y) is the coordinates of the current measured pixel, (x0, y0) is the center coordinates of the corner reflector "bright spot", The center coordinates of the calculated corner reflector "bright spot" are I(x, y), and the pixel value of the current pixel is I(x, y).

[0024] Further, the step S4 comprises the following steps:

[0025] (1) Calculate the azimuth, range and plane coordinate differences between the corner reflector point coordinates obtained in the step S3 and the corner reflector image point positions calculated by the distance Doppler model method;

[0026] (2) Take the average value of the plane coordinate differences of the plurality of corner reflector points on the SAR image as the true value, calculate the root mean square error (RSME) values of the azimuth, range and plane of the corner reflector points, and analyze the extraction results of the corner reflector points.

[0027] The root mean square error formula is:

[0028]

[0029] In the formula, n is the number of corner reflectors, x i represents the point position error value, f(x i ) represents the average value of the point position error. BRIEF DESCRIPTION OF DRAWINGS

[0030] The description of the content of the present application becomes obvious and easy to understand in combination with the following drawings, in which:

[0031] Figure 1 It is an analysis method flow chart of the adaptive weighted centroid method for extracting the corner reflector of the SAR image according to the present application.

[0032] Figure 2 It is a corner reflector image point subdivision interpolation flow chart according to the present application.

[0033] Figure 3 It is a distance-power index iterative weighted centroid model flow chart according to the present application. DETAILED DESCRIPTION

[0034] According toFigure 1 The step, the adaptive centroid weighting method of the application extracts the analysis method of the corner reflector point of the SAR image in detail.

[0035] Step 1: Determine the coarse position of the corner reflector point on the SAR image. It includes the following specific steps:

[0036] (1) In the area around the corner reflector image point obtained by the range-doppler model method, a window of a certain size is used to search for the brightest pixel value,

[0037] (2) Then, by artificial visual interpretation and analysis comparison of the surrounding ground features, the position of the maximum pixel value is determined as the coarse position of the corner reflector image point;

[0038] Step 2: Determine the window pixel information near the coarse position. It includes the following specific steps:

[0039] (1) Take the coarse position determined in step 1 as the coordinate, and select a certain pixel window size as the search window for high-precision corner reflector point coordinates;

[0040] (2) Record the pixel values and coordinates of each pixel obtained according to the search window in the extraction order;

[0041] Step 3: Construct the distance weighted centroid method based on the distance relationship between pixels to accurately extract the corner reflector point. It includes the following specific steps:

[0042] (1) According to the formula of the adaptive weighted centroid method, take the reciprocal of the distance as the weighting function, and calculate the position of the corner reflector point. The formula is as follows:

[0043]

[0044] In the formula, (x, y) is the coordinate of the current measured pixel, (x0, y0) is the center coordinate of the corner reflector "bright spot", is the calculated center coordinate of the corner reflector "bright spot", and I(x, y) is the pixel value of the current pixel;

[0045] (2) Since the gray value distribution of the corner reflector image point is Gaussian distribution, the gray value closer to the center is larger, so the distance weighted centroid method uses the square of the distance to increase the contribution of the pixels close to the center to the calculation of the centroid. The formula is as follows:

[0046]

[0047] In the formula, (x, y) is the coordinate of the current measured pixel, (x0, y0) is the center coordinate of the corner reflector "bright spot", is the calculated center coordinate of the corner reflector "bright spot", and I(x, y) is the pixel value of the current pixel.

[0048] Step 4: Precision evaluation. Including the following specific steps:

[0049] (1) Calculate the azimuth, range and plane coordinate differences between the corner reflector point coordinates obtained in step 3 and the corner reflector image point positions calculated by the range Doppler model method;

[0050] (2) Taking the average value of the plane coordinate differences of multiple corner reflector points on the SAR image as the true value, calculate the root mean square error (RSME) values of the azimuth, range and plane of the corner reflector points, and analyze the extraction results of the corner reflector points.

[0051] The root mean square error formula is:

[0052]

[0053] In the formula, n is the number of corner reflectors, x i represents the point error value, f(x i ) represents the average value of the point error value.

[0054] The application discloses a corner reflector image point fitting method based on an improved distance weighted centroid method. Through a fitting window of a certain size, a distance square weighted formula is constructed according to the distance relationship between the corner reflector image point pixels. The method has the characteristics of accurate extraction, and can provide a reference basis for the precision extraction of the corner reflector points on the SAR image.

[0055] The above only describes the preferred embodiments of the application and is not used to limit the application, and any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the application shall be included in the protection scope of the application.

Claims

1. An adaptive range-power exponential iterative weighted centroid method for corner reflector image point fitting, characterized in that, The method comprises the following steps: S1: determining the coarse position of the corner reflector image point: on the basis of the position of the corner reflector image point calculated by the range-doppler model method, the coarse position of the corner reflector image point on the SAR image is determined by artificial visual interpretation and analysis of the surrounding ground feature appearance; S2: adaptively judging the fitting window size: according to the pixel value at the coarse position of the corner reflector image point determined in S1, the window size occupied by the corner reflector image point on the SAR image is judged, and the fitting window size of each corner reflector image point is adaptively judged based on the window size; S3: determining the pixel information in the fitting window: the pixel values and corresponding coordinates of all pixels in the fitting window determined in S2 are sequentially extracted and recorded; S4: pixel subdivision in the fitting window: each pixel in the fitting window is subdivided, the subdivided pixel coordinates are recorded, and the pixel value of each subdivided pixel is calculated by using a certain mathematical function method; S5: constructing a range-power exponent iterative weighted centroid model: the step S5 comprises the following steps: (1) improving the formula based on the range-power exponent weighted centroid method, and the improvement mode is to take the reciprocal of the power of the distance as the weighting function to calculate the position of the corner reflector image point, and the formula is as follows: where (x, y) is the coordinate of the current pixel to be measured, (x0, y0) is the coordinate of the center of the corner reflector "bright spot", is the calculated coordinate of the center of the corner reflector "bright spot", I(x, y) is the pixel value of the current pixel; the calculated position coordinate is taken as the center position parameter of the weight function of the next iteration, and the "bright spot" is the point with the maximum pixel value searched by a window of a certain size; (2) iterative calculation: a certain threshold value is set, the difference between the center coordinates of the corner reflector "bright spot" calculated according to formula (1) and the coordinates of the maximum pixel value point is calculated, and the coordinate difference between the two is compared with the size of the threshold value; if it is greater than the set threshold value, the position of the corner reflector image point calculated by formula (1) is taken as the center position parameter of the next iteration value function and brought into formula (1) until the coordinate difference between the two is less than the set threshold value; the centroid position of the corner reflector image point is iteratively calculated until the accuracy of the obtained result meets the requirements, the iteration is ended, and the high-precision centroid position of the corner reflector image point is output; S6: accuracy evaluation: the position of the corner reflector image point calculated in S5 is compared with the position of the corner reflector image point calculated by the range-doppler model method, and the accuracy error of the corner reflector image point extracted by using the adaptive range-power exponent iterative weighted centroid method is analyzed.

2. The adaptive range-power exponential iterated weighted centroid method for corner reflector image point fitting according to claim 1, wherein The step S1 comprises the following steps: (1) a window of a certain size is used to search for the "bright spot" with the maximum pixel value near the corner reflector image point on the SAR image calculated by the range-doppler model method; (2) then, the maximum pixel value position is determined as the coarse position of the corner reflector image point by artificial visual interpretation and analysis of the surrounding ground feature appearance.

3. The adaptive range-power exponential iterated weighted centroid method for corner reflector image point fitting according to claim 1, wherein The step S2 comprises the following steps: (1) according to the pixel value of the corner reflector image point at the coarse position determined in step S1, the window range occupied by the corner reflector image point is judged; (2) based on the window range occupied by the corner reflector image point, the fitting window size of each corner reflector image point on the SAR image is adaptively judged.

4. The adaptive range-power exponential iterated weighted centroid method for corner reflector image point fitting of claim 1, wherein The step S3 comprises the following steps: (1) determining the fitting window size of the corner reflector image point; (2) sequentially extracting and recording the pixel values and corresponding coordinates of all pixels in the fitting window.

5. The corner reflector spot-fitting adaptive range-power exponential iterated weighted centroid method of claim 1, wherein The step S4 comprises the following steps: (1) subdividing each pixel in the fitting window, and recording the subdivided pixel coordinates; (2) using a certain mathematical function method to obtain the pixel value of each sub-pixel.

6. The corner reflector image point fitting adaptive range-power exponential iterated weighted centroid method of claim 1, wherein The step S6 comprises the following steps: (1) calculating the azimuth, range and plane coordinate differences between the corner reflector pixel coordinates obtained in the step S5 and the corner reflector pixel positions calculated by the range Doppler model method; (2) taking the average value of the plane coordinate differences of the plurality of corner reflector pixels on the SAR image as a true value, calculating the root mean square error RSME values of the azimuth, range and plane of the corner reflector pixels, and analyzing the extraction results of the corner reflector pixels, and the root mean square error formula is: where k is the number of corner reflectors, x i represents the point position error value, f(x i ) represents the average value of the point position error value.

Images