A CFAR ship target detection method based on local saliency enhancement

Abstract

The present invention relates to a CFAR ship target detection method based on local saliency enhancement, comprising: acquiring an original SAR image to be detected; performing local grayscale feature difference measurement to obtain a local grayscale feature contrast detection result; performing local texture feature difference measurement to obtain a local texture feature contrast detection result; fusing the local grayscale feature contrast detection result and the local texture feature contrast detection result to obtain a target enhancement saliency map, and then performing CFAR target detection based on generalized gamma distribution according to the target enhancement saliency map to obtain a ship target detection result. The present invention can well solve the shortcomings of low target detection accuracy and high false alarm rate caused by features such as speckle noise and side lobes; can significantly improve the target-clutter contrast TCR, has a good target enhancement effect and background suppression effect in a relatively complex multi-target environment, and significantly improves the detection performance of small ship targets.

Description

Technical Field

[0001] The invention relates to the technical field of radar remote sensing application, and in particular to a CFAR ship target detection method based on local saliency enhancement. Background Art

[0002] Synthetic Aperture Radar (SAR) is an active microwave remote sensor that can acquire high-resolution radar images. It has the advantages of working in all weather and all day, and is widely used in military fields and civil fields such as maritime traffic control and fishery management. Therefore, it is of great significance to study ship target detection in high-performance SAR images.

[0003] The traditional pixel-level constant false alarm rate (CFAR) algorithm is widely used. This detection method is based on the statistical modeling of sea clutter and uses the grayscale statistical characteristics between the target and the background to achieve target detection. However, the scattering of sea clutter is complex and statistical modeling is difficult, which will cause the detection performance of the detector to deteriorate. In addition, it is difficult to use a single grayscale feature in SAR images to distinguish between the target and the clutter background.

[0004] In recent years, the salient feature enhancement theory has been gradually applied to target detection tasks, which has great potential. In complex environments, the performance of ship target detection will be reduced due to speckle noise and uneven scattering of ship targets, especially for the detection of small-scale targets. At present, in view of the complex multi-target marine environment, the small target has weak scattering characteristics and only occupies a small number of pixels in the SAR image. There are still some difficulties in improving the performance of small target detection, and further research is still needed. Summary of the invention

[0005] The object of the present invention is to provide a CFAR ship target detection method based on local saliency enhancement, which can significantly improve TCR, suppress speckle noise, and has a higher detection rate and a lower false alarm rate in a relatively complex multi-target environment.

[0006] To achieve the above object, the present invention adopts the following technical solution: a CFAR ship target detection method based on local saliency enhancement, the method comprising the following steps in order:

[0007] (1) Obtain the original SAR image to be detected and design a diagonal-oblique sliding window;

[0008] (2) Perform local grayscale feature difference measurement on the original SAR image to be detected to obtain the local grayscale feature contrast detection result;

[0009] (3) Perform local texture feature difference measurement on the original SAR image to be detected to obtain the local texture feature contrast detection result;

[0010] (4) The local grayscale feature contrast detection results and the local texture feature contrast detection results are fused to obtain the target enhancement saliency map. Then, CFAR target detection based on the generalized gamma distribution is performed according to the target enhancement saliency map to obtain the ship target detection result.

[0011] The step (1) specifically refers to: selecting a multi-target scene SAR in the public data set HRSID as the original SAR image to be detected, designing a diagonal-diagonal sliding window, and the diagonal-diagonal sliding window is divided into three layers, the first layer is the target window T, the second layer is the protection window P, and the third layer is the four neighborhood background windows B1, B2, B3, and B4 on the diagonal-diagonal line.

[0012] The step (2) specifically refers to:

[0013] Based on the gray intensity contrast between the local window area blocks, the gray dissimilarity between the target block and its surrounding clutter background blocks is defined. The gray dissimilarity calculation formula is:

[0014]

[0015] Among them, μ T Represents the grayscale mean of the target block in the local sliding window area, Represents the grayscale mean of the four surrounding background blocks; when When , the greater the difference between the grayscale mean of the target block and the grayscale mean of the surrounding clutter background block, the greater the grayscale dissimilarity Lg; on the contrary, when When , no grayscale comparison is performed and the grayscale dissimilarity Lg is directly set to 1;

[0016] Combining the intensity characteristics of the target block and the grayscale dissimilarity between the target block and its surrounding clutter background blocks, the local grayscale feature contrast detection result C is obtained by local sliding window processing. g for:

[0017]

[0018] Where (p, q) is the center coordinate of the target block T, It represents the jth maximum grayscale value of the target block T, and N is the number of maximum grayscale values. When the grayscale mean of the target block is greater than the grayscale mean of the background block and the difference is greater, the grayscale enhancement effect on the target block is stronger. On the contrary, when the grayscale mean of the target block is less than the grayscale mean of the background block, the grayscale feature of the target block is not processed.

[0019] The step (3) specifically refers to:

[0020] The matrix pattern of the local texture structure described by LBP is expressed as:

[0021]

[0022] Among them, x c represents the grayscale value of the pixel at the center of the window, x(i,j) represents the grayscale value of any pixel in the window area, and L represents the size of the window; pixels with an S value of 1 are recorded as bright pixels, and pixels with an S value of 0 are recorded as dark pixels;

[0023] The texture intensity feature value is extracted for the non-uniform texture structure described by the local LBP. The calculation formula of the local texture intensity feature value is as follows:

[0024]

[0025] In the formula, m = |x(i,j)-x c |,i,j=1,2,...,L represents the absolute value of the grayscale difference between the central pixel of the L×L local area block and the remaining neighboring pixels, then |D| max and |D| min They represent the maximum and minimum values ​​of the absolute value of the grayscale difference, respectively, and P(m) represents the probability when the value is m;

[0026] The local texture feature contrast detection result C is obtained by sliding window calculation on the global SAR image. te for:

[0027]

[0028] In the formula, Lc T represents the texture intensity feature value of the target block in the center of the sliding window, Represents the texture intensity feature value of the surrounding clutter background block, It represents the grayscale mean of all bright pixels in the L×L target block area whose pixel intensity is greater than or equal to the central pixel intensity; It represents the grayscale mean of all dark pixels in the L×L target block whose pixel intensity is less than the central pixel intensity. Indicates the light-dark contrast of the target block.

[0029] The step (4) specifically refers to:

[0030] The calculation formula of integrating the local grayscale feature contrast detection result and the local texture feature contrast detection result is:

[0031] C s =C g ×C te

[0032] Among them, C g is the local grayscale feature contrast detection result, C te is the local texture feature contrast detection result, C s Enhance saliency map for the target;

[0033] The CFAR target detection based on the generalized gamma distribution is performed on the fused target enhancement saliency map, specifically:

[0034] Enhanced saliency map of target based on generalized gamma distribution model C s For CFAR ship target detection, the probability density function of the generalized gamma distribution is defined as:

[0035]

[0036] In the above formula, δ, v and k represent the scale parameter, power parameter and shape parameter respectively, and Γ(·) represents the gamma function;

[0037] The relationship between the detection threshold and the false alarm probability is:

[0038]

[0039] In the formula, Q Inv represents the inverse incomplete gamma function, P fa is the false alarm probability;

[0040] Based on the detection threshold Th calculated by the above formula, the saliency map C of the target is enhanced. s Perform target binary detection, that is, when C s When the value of any pixel in is greater than or equal to the detection threshold Th, the pixel is regarded as a target pixel, otherwise, it is regarded as a background pixel, and the ship target detection result C is obtained. detection It is expressed as:

[0041]

[0042] Among them, (i, j) represents the target enhanced saliency map C s At any pixel position in C detection When (i, j) = 1, the pixel represents the target pixel. detection When (i, j) = 0, the pixel represents a background pixel.

[0043] It can be seen from the above technical scheme that the beneficial effects of the present invention are: first, the present invention can well solve the shortcomings of low target detection accuracy and high false alarm rate caused by features such as speckle noise and side lobes; second, the present invention combines grayscale features and texture features for saliency enhancement processing, which can significantly improve the target-clutter contrast TCR, has good target enhancement effect and background suppression effect in a relatively complex multi-target environment, and significantly improves the detection performance of small ship targets. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] Figure 1 is a flow chart of the method of the present invention;

[0045] Figure 2 It is a diagonal-diagonal sliding window structure;

[0046] Figure 3 It is the original SAR image to be detected selected from the public dataset HRSID;

[0047] Figure 4 This is the result diagram of ship target detection. DETAILED DESCRIPTION

[0048] like Figure 1 As shown, a CFAR ship target detection method based on local saliency enhancement comprises the following steps in sequence:

[0049] (1) Obtain the original SAR image to be detected and design a diagonal-oblique sliding window;

[0050] (2) Perform local grayscale feature difference measurement on the original SAR image to be detected to obtain the local grayscale feature contrast detection result;

[0051] (3) Perform local texture feature difference measurement on the original SAR image to be detected to obtain the local texture feature contrast detection result;

[0052] (4) The local grayscale feature contrast detection results and the local texture feature contrast detection results are fused to obtain the target enhancement saliency map. Then, CFAR target detection based on the generalized gamma distribution is performed according to the target enhancement saliency map to obtain the ship target detection result.

[0053] The step (1) specifically refers to: selecting a multi-target scene SAR in the public data set HRSID as the original SAR image to be detected, such as Figure 3 As shown in , the image data comes from Sentinel-1B SAR satellite images, and the image size is 800×800 pixels and the image resolution is 3m. Considering the impact of background pixels and sidelobe features around the target on the detection performance, a diagonal-oblique diagonal sliding window is designed, as shown in Figure 2 As shown, the diagonal-diagonal sliding window is divided into three layers, the first layer is the target window, i.e., the target block T, the second layer is the protection window P, and the third layer is the four-neighborhood background windows B1, B2, B3, and B4 on the diagonal-diagonal line.

[0054] The step (2) specifically refers to:

[0055] Based on the gray intensity contrast between the local window area blocks, the gray dissimilarity between the target block and its surrounding clutter background blocks is defined. The gray dissimilarity calculation formula is:

[0056]

[0057] Among them, μ T Represents the grayscale mean of the target block in the local sliding window area, Represents the grayscale mean of the four surrounding background blocks; when When , the greater the difference between the grayscale mean of the target block and the grayscale mean of the surrounding clutter background block, the greater the grayscale dissimilarity Lg; on the contrary, when When , no grayscale comparison is performed and the grayscale dissimilarity Lg is directly set to 1; in order to avoid the denominator being zero, ε is generally taken as a very small positive number.

[0058] Combining the intensity characteristics of the target block and the grayscale dissimilarity between the target block and its surrounding clutter background blocks, the local grayscale feature contrast detection result C is obtained by local sliding window processing. g for:

[0059]

[0060] Where (p, q) is the center coordinate of the target block T, It represents the jth maximum grayscale value of the target block T, and N is the number of maximum grayscale values. When the grayscale mean of the target block is greater than the grayscale mean of the background block and the difference is greater, the grayscale enhancement effect on the target block is stronger. On the contrary, when the grayscale mean of the target block is less than the grayscale mean of the background block, the grayscale feature of the target block is not processed.

[0061] The step (3) specifically refers to: since it is difficult to achieve efficient detection of ship targets in SAR images by relying solely on grayscale features, especially for the detection of small ship targets with weak scattering and small volume, in order to more accurately extract the ship targets from the complex sea clutter background, the present invention mainly focuses on the texture difference between the target and the background, and further performs target enhancement and background suppression processing on the original SAR image to be detected.

[0062] Local Binary Pattern (LBP) is a description operator widely used to describe the texture features of image regions. Traditional LBP mainly uses the grayscale value of the central pixel in a 3×3 window as the threshold, and divides the remaining pixels in the neighborhood. The pixels with a grayscale value greater than the pixel are recorded as bright pixels, and the pixels with a grayscale value less than the pixel are recorded as dark pixels. Generally, the LBP pattern of a 3×3 neighborhood window of a certain pixel is a uniform pattern, but as the window size increases, the LBP pattern distribution will become uneven, and this non-uniform LBP pattern will provide more information representation in the window area composed of a certain pixel and its surrounding neighborhood pixels.

[0063] The matrix pattern of the local texture structure described by LBP is expressed as:

[0064]

[0065] Among them, x c represents the gray value of the pixel at the center of the window, x(i,j) represents the gray value of any pixel in the window area, L represents the size of the window, and is generally an odd integer greater than or equal to 3; pixels with an S value of 1 are recorded as bright pixels, and pixels with an S value of 0 are recorded as dark pixels;

[0066] The presence of speckle noise in the original SAR image and the uneven scattering of ship targets will make the relative difference in the groove intensity between the ship target and the clutter background larger, that is, the groove of the ship target is significantly stronger than that of the clutter background. Therefore, based on the non-uniform texture structure described by the LBP operator, the present invention first calculates the grayscale difference between the central pixel point of the local window area and any of its neighboring pixels, and completes the grayscale differential statistics of the local area. Then, the texture intensity feature value obtained by combining the regional grayscale differential statistics and the light-dark contrast of the region can better characterize the texture information of the local area.

[0067] The texture intensity feature value is extracted for the non-uniform texture structure described by the local LBP. The calculation formula of the local texture intensity feature value is as follows:

[0068]

[0069] In the formula, m = |x(i,j)-x c |,i,j=1,2,...,L represents the absolute value of the grayscale difference between the central pixel of the L×L local area block and the remaining neighboring pixels, then |D| max and |D| min They represent the maximum and minimum values ​​of the absolute value of the grayscale difference, respectively, and P(m) represents the probability when the value is m;

[0070] The local texture feature contrast detection result C is obtained by sliding window calculation on the global SAR image. te for:

[0071]

[0072] Where, Lc T represents the texture intensity feature value of the target block in the center of the sliding window, Represents the texture intensity feature value of the surrounding clutter background block, It represents the grayscale mean of all bright pixels in the L×L target block area whose pixel intensity is greater than or equal to the central pixel intensity; It represents the grayscale mean of all dark pixels in the L×L target block whose pixel intensity is less than the central pixel intensity. Indicates the light-dark contrast of the target block.

[0073] The step (4) specifically refers to:

[0074] The calculation formula of integrating the local grayscale feature contrast detection result and the local texture feature contrast detection result is:

[0075] C s =C g ×C te

[0076] Among them, C g is the local grayscale feature contrast detection result, C te is the local texture feature contrast detection result, C s Enhance saliency map for the target;

[0077] The CFAR target detection based on the generalized gamma distribution is performed on the fused target enhancement saliency map, specifically:

[0078] Enhanced saliency map of target based on generalized gamma distribution model C s For CFAR ship target detection, the probability density function of the generalized gamma distribution is defined as:

[0079]

[0080] In the above formula, δ, v and k represent the scale parameter, power parameter and shape parameter respectively, and Γ(·) represents the gamma function;

[0081] The relationship between the detection threshold and the false alarm probability is:

[0082]

[0083] In the formula, Q Inv represents the inverse incomplete gamma function, P fa is the false alarm probability;

[0084] Based on the detection threshold Th calculated by the above formula, the saliency map C of the target is enhanced. s Perform target binary detection, that is, when C s When the value of any pixel in is greater than or equal to the detection threshold Th, the pixel is regarded as a target pixel, otherwise, it is regarded as a background pixel, and the ship target detection result C is obtained. detection It is expressed as:

[0085]

[0086] Among them, (i, j) represents the target enhanced saliency map C s At any pixel position in detection When (i, j) = 1, the pixel represents the target pixel. detection When (i, j) = 0, the pixel represents a background pixel.

[0087] Enhance the saliency map C of the entire target s Perform pixel-by-pixel judgment to determine the target pixel and background pixel, thereby completing the binary detection of the target, such as Figure 4 As shown, it can be seen from the detection results that the invention can accurately detect ship targets and improve the target detection performance, especially the detection of small targets.

[0088] In summary, the present invention mainly utilizes the difference in local spatial distribution of grayscale features of target and background in SAR images and the difference in texture features to perform target enhancement and background suppression processing, and performs CFAR ship target detection under a certain false alarm probability based on the target enhancement saliency map, and the effectiveness of the invention is proved by a large number of experiments. The present invention can significantly improve TCR and has a low probability of missed detection and false alarm when performing target detection in a multi-target scene, can effectively suppress the influence of speckle noise, and significantly improve the detection performance of targets, especially small targets.

Claims

1. A CFAR ship target detection method based on local saliency enhancement, Features: The method comprises the following steps in order: (1) Obtain the original SAR image to be detected and design a diagonal-oblique sliding window; (2) Perform local grayscale feature difference measurement on the original SAR image to be detected to obtain the local grayscale feature contrast detection result; (3) Perform local texture feature difference measurement on the original SAR image to be detected to obtain the local texture feature contrast detection result; (4) The local grayscale feature contrast detection results and the local texture feature contrast detection results are integrated to obtain the target enhancement saliency map. Then, CFAR target detection based on the generalized gamma distribution is performed according to the target enhancement saliency map to obtain the ship target detection result. Step (1) specifically refers to: selecting a multi-target scene SAR in the public data set HRSID as the original SAR image to be detected, designing a diagonal-diagonal sliding window, wherein the diagonal-diagonal sliding window is divided into three layers, the first layer is the target window T, the second layer is the protection window P, and the third layer is the four neighborhood background windows B1, B2, B3, and B4 on the diagonal-diagonal line; Step (4) specifically refers to: The calculation formula of integrating the local grayscale feature contrast detection result and the local texture feature contrast detection result is: C s =C g ×C te Among them, C g is the local grayscale feature contrast detection result, C te is the local texture feature contrast detection result, C s Enhance saliency map for the target; The CFAR target detection based on the generalized gamma distribution is performed on the fused target enhancement saliency map, specifically: Enhanced saliency map of target based on generalized gamma distribution model C s For CFAR ship target detection, the probability density function of the generalized gamma distribution is defined as: In the above formula, δ, v and k represent the scale parameter, power parameter and shape parameter respectively, and Γ(·) represents the gamma function; The relationship between the detection threshold and the false alarm probability is: In the formula, Q Inv represents the inverse incomplete gamma function, P fa is the false alarm probability; Based on the detection threshold Th calculated by the above formula, the saliency map C of the target is enhanced. s Perform target binary detection, that is, when C s When the value of any pixel in is greater than or equal to the detection threshold Th, the pixel is regarded as a target pixel, otherwise, it is regarded as a background pixel, and the ship target detection result C is obtained. detection It is expressed as: Among them, (i, j) represents the target enhanced saliency map C s At any pixel position in detection When (i, j) = 1, the pixel represents the target pixel. detection When (i, j) = 0, the pixel represents a background pixel.

2. The CFAR ship target detection method based on local saliency enhancement according to claim 1, Features: Step (2) specifically refers to: Based on the gray intensity contrast between the local window area blocks, the gray dissimilarity between the target block and its surrounding clutter background blocks is defined. The gray dissimilarity calculation formula is: Among them, μ T Represents the grayscale mean of the target block in the local sliding window area, Represents the grayscale mean of the four surrounding background blocks; when When , the greater the difference between the grayscale mean of the target block and the grayscale mean of the surrounding clutter background block, the greater the grayscale dissimilarity Lg; on the contrary, when When , no grayscale comparison is performed and the grayscale dissimilarity Lg is directly set to 1; Combining the intensity characteristics of the target block and the grayscale dissimilarity between the target block and its surrounding clutter background blocks, the local grayscale feature contrast detection result C is obtained by local sliding window processing. g for: Where (p, q) is the center coordinate of the target block T, It represents the jth maximum grayscale value of the target block T, and N is the number of maximum grayscale values. When the grayscale mean of the target block is greater than the grayscale mean of the background block and the difference is greater, the grayscale enhancement effect on the target block is stronger. On the contrary, when the grayscale mean of the target block is less than the grayscale mean of the background block, the grayscale feature of the target block is not processed.

3. The CFAR ship target detection method based on local saliency enhancement according to claim 1, Features: Step (3) specifically refers to: The matrix pattern of the local texture structure described by LBP is expressed as: Among them, x c represents the grayscale value of the pixel at the center of the window, x(i,j) represents the grayscale value of any pixel in the window area, and L represents the size of the window; pixels with an S value of 1 are recorded as bright pixels, and pixels with an S value of 0 are recorded as dark pixels; The texture intensity feature value is extracted for the non-uniform texture structure described by the local LBP. The calculation formula of the local texture intensity feature value is as follows: In the formula, m = |x(i,j)-x c |,i,j=1,2,...,L represents the absolute value of the grayscale difference between the central pixel of the L×L local area block and the remaining neighboring pixels, then |D| max and |D| min They represent the maximum and minimum values ​​of the absolute value of the grayscale difference, respectively, and P(m) represents the probability when the value is m; The local texture feature contrast detection result C is obtained by sliding window calculation on the global SAR image. te for: In the formula, represents the texture intensity feature value of the target block in the center of the sliding window, Represents the texture intensity feature value of the surrounding clutter background block, It represents the grayscale mean of all bright pixels in the L×L target block area whose pixel intensity is greater than or equal to the central pixel intensity; It represents the grayscale mean of all dark pixels in the L×L target block whose pixel intensity is less than the central pixel intensity. Indicates the light-dark contrast of the target block.

Images