A weak and small target detection method, system and storage medium

Abstract

The application belongs to the technical field of target detection, and discloses a weak small target detection method, a system and a storage medium, which comprise the following steps: calculating the local contrast CM based on the gray distribution of a to-be-detected image at a current scale; fitting the local gray distribution of the to-be-detected image by using a two-dimensional discrete orthogonal polynomial; respectively calculating the partial derivatives of the fitted local gray distribution with respect to the rows and columns of the to-be-detected image to obtain the directional derivatives of the to-be-detected image in the vertical and horizontal directions, and then obtaining the directional derivative at any angle; extracting the features of each pixel point of the to-be-detected image on the directional derivative at the corresponding angle, and obtaining the directional derivative feature GM of the to-be-detected image after fusion; fusing the CM and the GM to obtain the saliency map of the to-be-detected image at the current scale; fusing the saliency maps at all scales to obtain the final saliency map S; and segmenting the saliency map S to locate the position of the target. The method has high accuracy and low false alarm rate for weak small targets.

Description

TECHNICAL FIELD

[0001] The present application belongs to the technical field of target detection, and more particularly, to a weak small target detection method and system and a storage medium. BACKGROUND

[0002] Infrared detection is widely used in non-cooperative target detection field in early stage due to its long imaging distance, all-day working, strong concealment and good anti-electromagnetic interference. The infrared weak small target detection technology aims to make full use of existing imaging system, combine with digital image processing technology, detect and locate the target of interest in the image, and give indication information.

[0003] In practical application, the weak small target detection technology still faces some problems and bottlenecks. First, the target is weak and small, usually only contains several or dozens of pixels, resulting in lack of enough shape, edge and other detail information of the target, which cannot design effective feature extraction method based on shape and texture features; then, the image background is complex, with more clutter, low signal-to-clutter ratio, and the target is easily submerged in complex background, plus the image produces isolated noise points, including fixed position oversaturated pixels, dead pixels and position random flicker pixels, which have similar features with the target, and are easily mistaken for targets in detection. The existing detection method based on background estimation has low detection accuracy when facing complex scenes or scenes with serious noise. SUMMARY

[0004] In view of the above defects or improvement needs of the prior art, the present application provides a weak small target detection method, system and storage medium, which aims to improve the accuracy of weak small target detection.

[0005] To achieve the above purpose, according to the first aspect of the present application, a weak small target detection method is provided, comprising:

[0006] calculating the local contrast CM based on gray distribution of the to-be-detected image at the current scale;

[0007] fitting the local gray distribution of the to-be-detected image by using two-dimensional discrete orthogonal polynomial; and taking partial derivative of the fitted local gray distribution with respect to the row and column of the to-be-detected image, to obtain the directional derivative of the to-be-detected image in the vertical direction and the horizontal direction, so as to calculate the directional derivative of the to-be-detected image at any angle;

[0008] extracting the feature of each pixel point of the to-be-detected image on the directional derivative of its corresponding angle, and fusing the features to obtain the directional derivative feature GM of the to-be-detected image;

[0009] fusing the local contrast CM and the directional derivative feature GM to obtain the saliency map S of the to-be-detected image at the current scale scale; and fusing the saliency maps under each scale to obtain a final saliency map S;

[0010] segmenting the saliency map S to locate a position of the target in the image to be detected.

[0011] Further, a local contrast CM based on a gray distribution of the image to be detected under the current scale is calculated, including:

[0012] traversing each pixel point of the image to be detected to construct a target region with the current pixel point of the image to be detected as a center; wherein the target region includes a target layer and a background layer B; the target layer is a square region with the current pixel point of the image to be detected as a center and a side length of the current scale; the background layer B is K sub-regions equal in size and adjacent to the target layer; and the target layer is composed of a target center layer C and a target attenuation layer A, the target center layer C is a region with a pixel size of 1x1 and the current pixel point as a center, and the remaining region in the target layer is the target attenuation layer A;

[0013] calculating local contrasts between the target attenuation layer A, the background layer B and the target center layer C in pairs;

[0014] calculating the local contrast CM based on the gray distribution of the image to be detected under the current scale: CM=C AB ×(C CA +C CB ); wherein C AB represents a local contrast between the target attenuation layer A and the background layer B; C CA represents a local contrast between the target center layer C and the target attenuation layer A; and C CB represents a local contrast between the target center layer C and the background layer B.

[0015] Further, the local contrasts between the target attenuation layer A, the background layer B and the target center layer C in pairs are calculated, including:

[0016]

[0017] wherein m A represents a mean value of the gray values of the target center layer C, m Bi represents a mean value of the gray values of the i-th sub-region of the background layer B, m C represents a mean value of the gray values of the target center layer C; max(a,b) represents a larger value of a and b, and min(a,b) represents a smaller value of a and b.

[0018] Further, a directional derivative of the image to be detected in an arbitrary angle is calculated, the arbitrary angle including 0°, 45°, 90° and 135°;

[0019] The features of each pixel point of the image to be detected on the directional derivative of the corresponding angle are extracted and fused to obtain the directional derivative feature GM of the image to be detected, including:

[0020] According to the position of each pixel point (r t , c t ) of the image to be detected, the L1 and L2 sub-regions are divided on the directional derivative of the corresponding angle, and the division method is:

[0021] L1 (0°) = {(r, c) | -s ≤ r-r t ≤ s, -2s ≤ c-c t ≤ 0}

[0022] L2 (0°) = {(r, c) | -s ≤ r-r t ≤ s, 0 ≤ c-c t ≤ 2s}

[0023] L1 (45°) = {(r, c) | -2s ≤ r-r t ≤ 0, -2s ≤ c-c t ≤ 0}

[0024] L2 (45°) = {(r, c) | 0 ≤ r-r t ≤ 2s, 0 ≤ c-c t ≤ 2s}

[0025] L1 (90°) = {(r, c) | -2s ≤ r-r t ≤ 0, -s ≤ c-c t ≤ s}

[0026] L2 (90°) = {(r, c) | 0 ≤ r-r t ≤ 2s, -s ≤ c-c t ≤ s}

[0027] L1 (135°) = {(r, c) | -2s ≤ r-r t ≤ 0, 0 ≤ c-c t ≤ 2s}

[0028] L2 (135°) = {(r, c) | 0 ≤ r-r t ≤ 2s, -2s ≤ c-c t ≤ 0}

[0029] Wherein, L1 (0°) and L2 (0°) , L1 (45°) and L2 (45°) , L1(90°) and L2 (90°) , L1 (135°) and L2 (135°) respectively correspond to L1 and L2 sub-regions divided on directional derivatives at 0°, 45°, 90°, 135° directions; r and c respectively are rows and columns of the image to be detected; l is the side length of the divided L1 and L2 regions, denotes a floor operation;

[0030] calculate the features GM α of L1 and L2 sub-regions divided on each directional derivative α : GM α = mean(L1-L2) x std(L1-L2); wherein GM t denotes a feature of each pixel point (r t , c t ) of the image to be detected on the directional derivative at an angle of a, a takes 0°, 45°, 90°, 135°, mean(*) denotes to take an average, and std(*) denotes to take a standard deviation;

[0031] After fusing GM α , the directional derivative feature GM of the image to be detected is obtained.

[0032] Further, a local gray scale distribution of the image to be detected is fitted by using two-dimensional discrete orthogonal polynomials, comprising:

[0033] select a discrete symmetric set R, and based on the orthogonality of polynomials, obtain two-dimensional discrete orthogonal polynomials P i (r, c), i = 1, 2,..., 2k corresponding to the discrete symmetric set R; wherein k is the number of elements in the discrete symmetric set R, and r and c respectively are rows and columns of the image to be detected;

[0034] calculate the coefficients k i of each two-dimensional discrete orthogonal polynomial:

[0035]

[0036] wherein W i is a weight matrix corresponding to the two-dimensional discrete orthogonal polynomial, and I(r, c) denotes the local gray scale distribution of the image to be detected;

[0037] multiply and accumulate each two-dimensional discrete orthogonal polynomial P i (r, c) and its corresponding coefficient k i , and obtain the fitted local gray scale distribution.

[0038] Further, the number of elements in the discrete symmetry set R is k=5, and the corresponding discrete symmetry set R is: R={ -2, -1, 0, 1, 2}.

[0039] Further, the local contrast CM and the directional derivative feature GM are fused, and the fusion manner is:

[0040] S scale = GM x CM

[0041] Wherein, S scale is a saliency map of the to-be-detected image at the current scale.

[0042] Further, the saliency map S is segmented by using an adaptive threshold segmentation manner, and a region with a value of 1 in the segmented saliency map is a position of the target in the to-be-detected image.

[0043] According to a second aspect of the present application, a weak small target detection system is provided, comprising a computer readable storage medium and a processor;

[0044] The computer readable storage medium is used to store executable instructions;

[0045] The processor is used to read the executable instructions stored in the computer readable storage medium to execute the weak small target detection method according to any one of the first aspect.

[0046] According to a third aspect of the present application, a computer readable storage medium is provided, and the computer readable storage medium stores a computer program, and the program is executed by a processor to implement the weak small target detection method according to any one of the first aspect.

[0047] Overall, the above technical solutions conceived by the present application can achieve the following beneficial effects:

[0048] (1) The weak small target detection method of the present application not only extracts the local contrast CM based on the gray distribution of the to-be-detected image at different scales, but also mines the directional derivative feature GM of the to-be-detected image to enhance the local contrast of the image. The target has obvious gray difference in each direction, and further has obvious directional derivative difference, so as to calculate the directional derivative feature GM of the to-be-detected image. The local contrast CM and the directional derivative feature GM are fused to enhance the local contrast. Therefore, the saliency map based on the fusion of the local contrast CM and the directional derivative feature GM is segmented, which can improve the accuracy of weak small target detection. At the same time, simulation experiments also prove that the method of the present application has high accuracy and low false alarm rate.

[0049] (2) As preferred, the application provides a local contrast CM calculation method based on gray distribution, different from the traditional two-layer model, the application adopts a three-layer model, divides the target region corresponding to each pixel point of the image to be detected into a background layer B, a target attenuation layer A and a target center layer C, that is, further divides the target layer into the target attenuation layer A and the 1x1 target center layer C, extracts the local contrast feature of the 1x1 target center layer and participates in the subsequent target detection judgment, avoids the problem that the existing two-layer model structure directly filters out the 1x1 region feature, which may lose the target feature, and can further improve the detection accuracy.

[0050] In addition, considering that the difference between the target attenuation layer A and the background layer B plays an absolute role, therefore, when the local contrast between the regions is fused to generate the final image local contrast CM, the designed fusion method can highlight the feature difference between the target attenuation layer A and the background layer B, and enhance the local feature of the target.

[0051] (3) As preferred, considering the difference between the target and the background, the real target generally conforms to the two-dimensional Gaussian distribution, while the noise does not have this distribution characteristic, therefore, when calculating the local contrast between the regions, not only the difference between the two regions is considered, but also the ratio between the two regions, and then the difference and the ratio between the regions are fused as the final image local contrast CM, in this way, the difference between the target and the background can be further highlighted, and the local feature of the target is further enhanced. At the same time, using the gray mean value of the region to represent the feature of the corresponding region also greatly reduces the complexity of the calculation.

[0052] (4) Further, when extracting the feature on each direction derivative, it is fully considered that the target to be detected should have obvious features in each direction derivative, a sub-region division method is designed, each direction derivative feature is extracted and fused, and the direction derivative feature obtained in this way can enhance the target while avoiding the influence of the background structure noise with obvious directionality on the detection.

[0053] (5) As preferred, the number of elements in the discrete symmetric set R is selected as 5, which can balance the calculation complexity and accuracy. BRIEF DESCRIPTION OF DRAWINGS

[0054] Figure 1 A weak small target detection method in the embodiment of the application.

[0055] Figure 2 A three-layer model for dividing the target region in the embodiment of the application.

[0056] Figure 3 Visualization results of the direction derivatives of the four directions of the ideal infrared weak small target.

[0057] Figure 4 L1 and L2 division manners corresponding to each directional derivative in the embodiment of the present application.

[0058] Figure 5 Three-dimensional distribution diagrams of local contrast CM, directional derivative feature GM and saliency map S corresponding to different images and corresponding segmentation results in the embodiment of the present application. DETAILED DESCRIPTION

[0059] In order to make the objectives, technical solutions and advantages of the present application clearer, the present application will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application. In addition, the technical features involved in each embodiment of the present application described below can be combined with each other as long as they do not conflict with each other.

[0060] Embodiment 1

[0061] As shown in the figure, the embodiment of the present application provides a weak and small target detection method, mainly including: Figure 1

[0062] S1, calculating the local contrast CM of the to-be-detected image at the current scale based on the gray distribution;

[0063] S2, fitting the local gray distribution of the to-be-detected image by using discrete orthogonal polynomials; and taking partial derivatives of the fitted local gray distribution of the to-be-detected image with respect to the rows and columns of the to-be-detected image, to obtain the directional derivatives of the to-be-detected image in the vertical direction and the horizontal direction; and calculating the directional derivative of the to-be-detected image at any angle based on the directional derivatives in the vertical direction and the horizontal direction;

[0064] S3, extracting the feature of each pixel point of the to-be-detected image on the directional derivative at its corresponding angle, and fusing to obtain the directional derivative feature GM of the to-be-detected image;

[0065] S4, fusing the local contrast CM and the directional derivative feature GM to obtain the saliency map S of the to-be-detected image at the current scale scale ; and fusing the saliency maps S at each scale scale as the final saliency map S;

[0066] S5, segmenting the final saliency map S to locate the target position.

[0067] ​It should be noted that, in the embodiments of the present application, in order to facilitate the description of the specific implementation manner or the preferred implementation manner of each step (such as in order to facilitate the following S1.1, S1.2, …, etc. numbering for the steps in S1), the S1-S5 sequential description manner is adopted, but in other embodiments, it can not be necessarily limited to such execution order, for example, S2 and S3 can also be executed before S1.

[0068] As one of the implementation manners, the existing local contrast algorithm can be used to calculate the local contrast CM based on the gray distribution of the to-be-detected image at the current scale, and in the embodiments of the present application, in order to improve the effect of target enhancement and reduce the calculation complexity, a specific calculation method for calculating the local contrast CM based on the gray distribution of the to-be-detected image at the current scale is provided, mainly including:

[0069] S1.1, each pixel point of the to-be-detected image is traversed by using a sliding window, and a target region is constructed with the current pixel point of the to-be-detected image as the center; wherein the target region includes a target layer and a background layer B; a square region with the current scale as the side length is taken as the target layer of the current target region with the current pixel point of the to-be-detected image as the center, and K sub-regions equal in size to the target layer and adjacent to the target layer are taken as the background layer of the current target region; and the target layer is composed of a target center layer C and a target attenuation layer A, the target center layer C is a region with the current pixel point as the center and the pixel size of 1*1, and the remaining regions in the target layer are the target attenuation layer A. As shown in the figure, Figure 2 in the embodiments of the present application, K is 8, that is, the background layer is divided into 8 sub-regions equal in size to the target layer, and numbered as 1-8.

[0070] S1.2, the local contrast between the target attenuation layer A, the background layer B and the target center layer C is calculated; in the embodiments of the present application, a local contrast calculation manner based on the region mean value and the ratio is provided, including:

[0071]

[0072] In formula (1), C represents the target center layer, I j represents the gray value of each pixel in C, N C represents the number of pixels in C, in the embodiments of the present application, N C =1, m C represents the mean value of the gray value of the target center layer C. In the embodiments of the present application, K is 8.

[0073] In formula (2), A represents the target center layer, I j represents the gray value of each pixel in A, N A represents the number of pixels in A, and under the current scale s, N A=s 2 -N C , m A represents the mean value of the gray value of the target center layer C.

[0074] B i represents the i-th sub-region of the background layer, I j represents the gray value of each pixel in B i , N Bi represents the number of pixels in B i , N Bi = s 2 , m Bi represents the mean value of the gray value of the i-th sub-region of the background layer.

[0075] max(a, b) in formula (4) represents taking the larger value of a and b, C CA represents the local contrast of the target center layer C and the target attenuation layer A.

[0076] min(a, b) in formula (5) represents taking the smaller value of a and b, C CB represents the local contrast of the target center layer C and the background layer B.

[0077] C AB represents the local contrast of the target attenuation layer A and the background layer B.

[0078] S1.3, calculate the gray distribution-based local contrast CM of the image to be detected at the current scale, and the calculation method is:

[0079] CM = C AB × (C CA + C CB ) (7)

[0080] CM in formula (7) represents the gray distribution-based local contrast of the image to be detected at the current scale.

[0081] The calculation method of the gray distribution-based local contrast CM provided in the embodiment of the application is different from the traditional two-layer model. In the embodiment of the application, a three-layer model is used, and the target region corresponding to each pixel point of the image to be detected is divided into a background layer B, a target attenuation layer A and a target center layer C. That is, the target layer is further divided into a target attenuation layer A and a 1*1 target center layer C. Considering that the image bad element is similar to the target pixel value, in order to avoid the interference of the image bad element on the target detection, the traditional method directly filters out the characteristics of the 1*1 region, which may lose the characteristics of the target. In the embodiment of the application, through the setting of the 1*1 target center layer and the subsequent feature extraction and participation in the subsequent target detection judgment, the detection accuracy can be improved.

[0082] In this embodiment of the present invention, considering the differences between the target and background, real targets generally follow a quasi-two-dimensional Gaussian distribution, while noise often lacks this distribution characteristic. Therefore, when calculating the local contrast between regions, not only the difference between the two regions is considered, but also the ratio between the two regions. These differences and ratios are then combined to form the final image local contrast CM. This further highlights the difference between the target and background. Furthermore, the grayscale mean of the target attenuation layer A, background layer B, and target center layer C is calculated using a regional mean method to extract features from the corresponding regions, significantly reducing computational complexity.

[0083] In addition, considering that the difference between the target attenuation layer A and the background layer B plays an absolute role, when fusion is performed based on the local contrast between regions to generate the final image local contrast CM, the designed fusion method can highlight the characteristic differences between the target attenuation layer A and the background layer B, further enhancing the local features of the target.

[0084] S2 includes:

[0085] S2.1. Select a discrete symmetric set R and solve the discrete orthogonal polynomial through iterative calculation. The calculation rules are as follows:

[0086]

[0087] Where, P n (r) and P k (r) represents polynomials of degree n and k respectively, k is the number of elements in the discrete symmetric set R, n = 0, 1..., k-1. The discrete symmetric set R can be represented by In the embodiment of the present invention, in order to balance the computational complexity and accuracy, k=5 is selected, that is, the corresponding discrete symmetric set R={-2,-1,0,1,2}.

[0088] According to the calculation rules of formula (8), the orthogonal polynomial corresponding to the discrete symmetric set R can be obtained as:

[0089] P n (r)={f(r 0 ),f(r 1 )...,f(r k-1 )} (9)

[0090] In the embodiment of the present invention, the orthogonal polynomial corresponding to the discrete symmetric set R={-2,-1,0,1,2} is:

[0091] P n (r) = {1,r,r 2 -2,r 3 -17 / 5r,r 4 -31 / 7r2 +72 / 35}

[0092] Extending the one-dimensional discrete orthogonal polynomial to two dimensions, the corresponding discrete orthogonal polynomial is:

[0093] P n (r,c)={f(r m ,c n )|m=0,1...,k-1,n=0,1...,k-1} (10)

[0094] In the embodiment of the present invention, since there are only 5 terms in one dimension, the corresponding two-dimensional discrete orthogonal polynomial only takes 10 terms of lower degree, specifically:

[0095] P n (r,c)={1,r,c,r 2 -2,rc,c 2 -2,r 3 -17r / 5,(r 2 -2)c,r(c 2 -2),c 3 -17c / 5}

[0096] S2.2, according to the discrete orthogonal polynomials in the above formula (10), each polynomial is multiplied by the corresponding coefficient k i And the local grayscale distribution I(r,c) of the image to be detected can be fitted by accumulating, where (r,c) represent the rows and columns of the image to be detected, respectively. Specifically, it can be expressed as:

[0097]

[0098] In the formula, since the number of elements in the discrete symmetric set is k, after extending from one dimension to two dimensions, the number of polynomials is 2k, so the corresponding coefficient k i The number is 2k.

[0099] S2.3, according to the two-dimensional discrete orthogonal polynomial P i (r,c) construct the weight matrix W i , to calculate the coefficient k of each polynomial i , and its calculation rules are:

[0100]

[0101] Where * represents the convolution operation.

[0102] Get the coefficient k of each polynomial i Then, according to formula (11), the local grayscale distribution of the image to be detected can be obtained by fitting discrete orthogonal polynomials.

[0103] S2.4. Take the partial derivatives of the local grayscale distribution of the image to be detected with respect to the rows and columns (r, c) of the image using discrete orthogonal polynomial fitting, and obtain the directional derivative of the image to be detected in the vertical direction. and the directional derivative of the image to be detected in the horizontal direction In the embodiment of the present invention, combined with the selected discrete symmetry set, the partial derivative of the image with respect to the row and column (r, c) can be expressed as:

[0104]

[0105] It can be observed that there are only 6 polynomial coefficients involved in the formula, and the corresponding weight matrix is:

[0106]

[0107] W3=W2' (19)

[0108] W 10 =W7' (20)

[0109] W9=W8' (21)

[0110] Among them, W2', W7', and W8' respectively represent the transpose of W2, W7, and W8.

[0111] S2.5. Calculate the directional derivative of the image to be detected at any angle based on the directional derivatives in the vertical and horizontal directions. The calculation formula is:

[0112]

[0113] Where α is an arbitrary angle, I' α Represents the directional derivative of the image to be detected at angle α. In the embodiment of the present invention, the directional derivative of the image to be detected at {0°, 45°, 90°, 135°} is calculated. Figure 3 As shown in the figure, the directional derivatives of the four angles of 0°, 45°, 90°, and 135° are obtained after the above steps of processing of the ideal infrared weak target, including a two-dimensional grayscale visualization map and a three-dimensional distribution map; Figure 3 The two pictures in the first row are the two-dimensional grayscale visualization and three-dimensional distribution of ideal infrared weak targets; the four pictures in the second row are the two-dimensional grayscale visualization of the directional derivatives at angles of 0°, 45°, 90°, and 135°; the four pictures in the third row are the three-dimensional distribution of the directional derivatives at angles of 0°, 45°, 90°, and 135°.

[0114] Specifically, in S3, an embodiment of the present invention provides a method for extracting each pixel point (r t , c t) on its corresponding directional derivatives, including:

[0115] S3.1, based on the directional derivatives of the image to be detected at 0°, 45°, 90°, and 135° obtained in S2, according to each pixel point (r t , c t ) position, divide the L1 and L2 sub-regions on the four directional derivatives, such as Figure 4 As shown, the division method is:

[0116]

[0117] Among them, formula (23) represents the L1, L2 division rule of the 0° direction derivative, L1 (0°) and L2 (0°) represents the L1 and L2 sub-regions divided on the 0° direction derivative; Formula (24) represents the L1 and L2 division rule of the 45° direction derivative, L1 (45°) and L2 (45°) represents the L1, L2 sub-regions divided on the 45° directional derivative; Formula (25) represents the L1, L2 division rule of the 90° directional derivative, L1 (90°) and L2 (90°) represents the L1, L2 sub-regions divided on the 90° directional derivative; Formula (26) represents the L1, L2 division rule of the 135° directional derivative, L1 (135°) and L2 (135°) Represents the L1 and L2 sub-regions divided on the 135° directional derivative; l is the side length of the divided L1 and L2 regions, where l is consistent with the current scale; Indicates a floor operation.

[0118] S3.2. Calculate the feature GM of the L1 and L2 sub-regions divided on each directional derivative α :

[0119] GM α =mean(L1-L2)×std(L1-L2) (28)

[0120] Where, GM α It represents the features of the L1 and L2 sub-regions on the directional derivative at an angle of α, that is, the component of the directional derivative feature GM of the image to be detected at an angle of α. mean(*) represents the average value, and std(*) represents the standard deviation.

[0121] Based on the components of each angle, the directional derivative feature GM of the image to be detected is calculated by multiplication:

[0122] GM=ΠGM α (29)

[0123] In the embodiment of the present application, when extracting the features on each direction derivative, it is fully considered that the target to be detected should have obvious features in each direction derivative, the sub-region division mode is designed, each direction derivative feature is extracted and fused, and thus the direction derivative feature obtained can enhance the target while avoiding the influence of the background structure noise with obvious directionality on the detection.

[0124] Specifically, in S4, the following steps are included:

[0125] The local contrast CM and the direction derivative feature GM are fused to obtain the saliency map S of the image to be detected at the current scale scale :

[0126] S scale = GM x CM (30)

[0127] As a preferred, in the embodiment of the present application, the saliency maps S scale at each scale are fused by using the maximum fusion mode to obtain the final saliency map S. In the embodiment of the present application, the saliency maps S scale at each scale are recorded as and the formula for fusing by using the maximum fusion mode is:

[0128]

[0129] In the formula, Q represents a target scale set, t i represents each target scale, represents the saliency map at the target scale t i , and max (*) represents the maximum value. S represents the final generated saliency map.

[0130] As a preferred, in S5, the adaptive threshold segmentation mode is used to segment the final saliency map S to locate the target position, and the steps include:

[0131] The adaptive segmentation threshold T is obtained according to the final saliency map S, and the calculation rule is:

[0132] T = μ + k' x σ (31)

[0133] In the formula, T represents the segmentation threshold, μ represents the mean value of the saliency map S, σ represents the standard deviation of the saliency map, and k' represents an empirical parameter. In the embodiment of the present application, the value range is 3-14.

[0134] The saliency map S is binarized according to the segmentation threshold, and the calculation rule is:

[0135]

[0136] The region with the saliency map value of 1 is the target position.

[0137] As shown in the figure, the detection result graphs on different images from left to right are an infrared original image to be detected, CM, GM, S and a threshold segmentation graph in the embodiment of the application, it can be found that the application shows high accuracy and low false alarm rate for the problem of weak and small target detection under complex background. The application has excellent background suppression ability and target enhancement ability. Figure 5

[0138] The weak and small target detection method of the application, in addition to extracting the local contrast CM based on the gray distribution of the image to be detected under different scales, also mines the directional derivative feature GM of the image to be detected to enhance the local contrast of the image, uses the obvious gray difference of the target in each direction, and further uses the obvious directional derivative difference to calculate the directional derivative feature GM of the image to be detected, fuses the local contrast CM and the directional derivative feature GM to enhance the local contrast, and therefore, the accuracy of weak and small target detection can be improved by segmenting the saliency graph fused based on the local contrast CM and the directional derivative feature GM. At the same time, simulation experiments also prove that the method has high accuracy and low false alarm rate.

[0139] Moreover, the existing method based on principal component analysis often needs to be iteratively calculated, and the algorithm has large calculation complexity and poor real-time performance, the method of the application can greatly reduce the complexity of the algorithm and has good real-time performance by using the regional mean to calculate the gray mean of the target attenuation layer A, the background layer B and the target center layer C to extract the features of the corresponding regions, and using a discrete symmetric set k=5.

[0140] Compared with the method based on deep learning, the method of the application has stronger universality.

[0141] In summary, the method of the application can effectively improve the detection rate of weak and small targets and reduce the false alarm rate under low calculation complexity when there is a complex background in the image, for example, containing a large number of clouds, and the image contains non-uniform components, for example, there are blind elements in the image, and the calculation real-time performance and accuracy are improved.

[0142] Embodiment 2

[0143] The weak and small target detection system provided in the embodiment of the application comprises a computer readable storage medium and a processor.

[0144] The computer readable storage medium is used to store executable instructions.

[0145] ​The processor is configured to read executable instructions stored in the computer readable storage medium to implement the weak small target detection method in embodiment 1. For related technical solutions, refer to the corresponding description in embodiment 1, which will not be repeated here.

[0146] Embodiment 3

[0147] The embodiment of the present application provides a computer readable storage medium, which stores a computer program. The program is executed by a processor to implement the weak small target detection method in embodiment 1. For related technical solutions, refer to the corresponding description in embodiment 1, which will not be repeated here.

[0148] Those skilled in the art can easily understand that the above description is only a preferred embodiment of the present application, and is not intended to limit the present application. Any modification, equivalent replacement and improvement made within the spirit and principle of the present application shall be included in the protection scope of the present application.

Claims

1. A small target detection method, characterized in that: include: Calculate the local contrast CM based on the grayscale distribution of the image to be detected at the current scale; A two-dimensional discrete orthogonal polynomial is used to fit the local grayscale distribution of the image to be detected; and partial derivatives of the fitted local grayscale distribution with respect to the rows and columns of the image to be detected are respectively calculated to obtain directional derivatives of the image to be detected in the vertical and horizontal directions, so as to calculate the directional derivative of the image to be detected at any angle; Extract the features of each pixel point of the image to be detected on the directional derivative of its corresponding angle, and after fusion, obtain the directional derivative feature GM of the image to be detected; The local contrast CM and the directional derivative feature GM are fused to obtain the saliency map of the image to be detected at the current scale. ; And fuse the saliency maps at each scale as the final saliency map S; Segmenting the saliency map S to locate the position of the target in the image to be detected; Calculate the local contrast CM based on the grayscale distribution of the image to be detected at the current scale, including: Traverse each pixel of the image to be detected and construct a target region with the current pixel of the image to be detected as the center; wherein the target region includes a target layer and a background layer B; the target layer is a square region with the current pixel of the image to be detected as the center and the current scale as the side length; the background layer B is K sub-regions of equal size and adjacent to the target layer; and the target layer is composed of a target center layer C and a target attenuation layer A. The target center layer C is a region with a pixel size of 1×1 with the current pixel as the center, and the remaining area in the target layer is the target attenuation layer A; Calculate the local contrast between the target attenuation layer A, the background layer B and the target center layer C; Calculate the local contrast CM based on the grayscale distribution of the image to be detected at the current scale: ;in, Represents the local contrast between the target attenuation layer A and the background layer B; Represents the local contrast between the target center layer C and the target attenuation layer A; Represents the target center layer Local contrast with background layer B; Calculate the local contrast between the target attenuation layer A, the background layer B, and the target center layer C, including: in, Represents the target center layer The mean of the grayscale values, Represents background layer B The mean gray value of the sub-regions, Represents the target center layer The mean of grayscale values; Indicates taking The larger value of Indicates taking The smaller value in .

2. The small target detection method according to claim 1, wherein: Calculate the directional derivative of the image to be detected at any angle, the arbitrary angle includes ; Extract the features of each pixel point of the image to be detected on the directional derivative of its corresponding angle, and after fusion, obtain the directional derivative feature GM of the image to be detected, including: According to each pixel of the image to be detected ( , ) position, divided on the directional derivative of its corresponding angle and Sub-areas are divided into: in, and 、 and 、 and 、 and Respectively expressed in Partitioned on directional derivatives and sub-region; and are the rows and columns of the image to be detected respectively; For the division and The side length of the region, Indicates a round-down operation; Calculate the division of the derivatives of each party and Characteristics of sub-regions : ;in, Represents each pixel of the image to be detected ( , )exist Characteristics of the directional derivative of the angle, Pick , Indicates finding the average value, Indicates the standard deviation. right After fusion, the directional derivative feature GM of the image to be detected is obtained.

3. The small target detection method according to claim 1, wherein: The local grayscale distribution of the image to be detected is fitted using a two-dimensional discrete orthogonal polynomial, including: Select discrete symmetric set , based on the orthogonality of polynomials, we get the discrete symmetric set The corresponding two-dimensional discrete orthogonal polynomials , ;in, For the discrete symmetric set The number of elements in , and ; and are the rows and columns of the image to be detected respectively; Compute the coefficients of each 2D discrete orthogonal polynomial : in, is the weight matrix corresponding to the two-dimensional discrete orthogonal polynomial, Represents the local grayscale distribution of the image to be detected; Each two-dimensional discrete orthogonal polynomial The corresponding coefficient After multiplication and accumulation, the fitted local grayscale distribution is obtained.

4. The small target detection method according to claim 3, wherein: The discrete symmetric set The number of elements in , the corresponding discrete symmetric set for: .

5. The small target detection method according to claim 1, wherein: The local contrast CM and the directional derivative feature GM are fused in the following manner: in, is the saliency map of the image to be detected at the current scale.

6. The small target detection method according to any one of claims 1 to 5, characterized in that: The saliency map S is segmented by using an adaptive threshold segmentation method, and the area with a value of 1 in the segmented saliency map is the position of the target in the image to be detected.

7. A small target detection system, characterized in that: comprising a computer-readable storage medium and a processor; The computer-readable storage medium is used to store executable instructions; The processor is configured to read the executable instructions stored in the computer-readable storage medium to execute the small target detection method according to any one of claims 1 to 6.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the small target detection method according to any one of claims 1 to 6 is implemented.

Images