An infrared small target detection method based on gray scale difference and information entropy

Abstract

The application discloses an infrared small target detection method based on gray scale difference and information entropy, and belongs to the technical field of infrared image detection. The application solves the problems of insufficient suppression ability of a selected feature to a complex background and low detection accuracy of infrared small target detection based on the selected feature. The extracted gradient module feature can make the target and other regions obviously different. The target region can be enhanced by extracting a direction angle feature and combining filtering, background region suppression is realized, and the local information entropy is used as a weight factor, which can effectively supplement the information loss caused in the gradient domain feature extraction process. Finally, by integrating the filtered direction angle feature map, the processed gradient module feature map and the entropy weight feature map, the accuracy of the infrared small target detection can be improved, and the universality of multiple use environments is improved. The method can be applied to infrared small target detection.

Description

TECHNICAL FIELD

[0001] The present application belongs to the technical field of infrared image detection, and particularly relates to an infrared small target detection method based on gray scale difference and information entropy. BACKGROUND

[0002] With the continuous development of infrared technology, the use scenarios of various types of infrared detection devices are more diverse. As a passive detection method, infrared detection has stronger concealment. The infrared detection system is less affected by light and has strong penetration capability, and can implement all-weather reconnaissance and early warning detection, and can cope with the influence of smoke, fog, haze and other adverse weather. Infrared technology is widely used in civil and military fields, including air-sea target monitoring, reconnaissance and early warning, search and rescue, etc. When the distance between the target to be detected and the infrared detection system is far, the target usually exists in the form of a small target in the imaging picture, so it is necessary to accurately and quickly detect the infrared small target image. The international organization SPIE defines that the target with an area size ratio of not more than 0.12% of the whole image is classified as a small target. On the one hand, the infrared small target has the characteristics of weak energy, small size, few pixels, and unobvious texture features; on the other hand, due to the different application scenarios, the influence of complex background and noise interference makes the infrared small target detection more difficult.

[0003] The existing infrared small target detection research mainly focuses on single frame image and multi-frame sequence image. At present, multi-frame image detection mainly includes two types. One is the correlation verification type, that is, first, a single frame image is used to obtain several possible targets, and then the continuity of the target motion trajectory is used to verify and determine in the multi-frame image to extract the motion trajectory of the real target. The second is the direct solution method, that is, the gray scale change information in the time domain is used to directly extract the target in the multi-frame image. In the face of high-speed and flexible motion devices, the target runs fast, the background changes rapidly, the system continuous detection is difficult, and it is difficult to obtain complete sequence level infrared detection image. On the other hand, the high-precision single frame image detection algorithm is the basis and premise of multi-frame image target detection. However, the infrared small target detection method of single frame image is based on artificial feature selection, mainly focusing on image space-frequency characteristics, local contrast characteristics, low-rank sparse characteristics to carry out detection task. For example, the patent application with publication number CN118521768A discloses an "infrared small target detection method based on non-window structure local contrast measurement", which detects the target by using the gradient characteristics of the image through mean filtering and gradient operator, but it is difficult to filter out pixel-level highlight clutter and strong edges, so there is still a significant deficiency in the suppression ability of complex background. The patent application with publication number CN118397302A discloses an "infrared small target detection method based on differential directional mixed sparse constraint combined with significant coherence enhancement", which generates a significant coherence consistency factor by processing multiple infrared images and then detects, but the model relies too much on the time and space information in the infrared image data, which is easy to cause the problems of background reduction and target loss.

[0004] In summary, the features selected by the existing method still have the problem of insufficient suppression ability for complex background. When the infrared small target detection is based on the features selected by the existing method, there is still the problem of low detection accuracy. Therefore, it is necessary to propose a new infrared image feature extraction method to suppress complex background and improve detection accuracy. SUMMARY

[0005] The purpose of the present application is to solve the problem of insufficient suppression ability of the features selected by the existing method for complex background, and the problem of low detection accuracy obtained by the infrared small target detection based on the features selected by the existing method. A method for detecting infrared small targets based on gray scale difference and information entropy is proposed.

[0006] The technical scheme adopted by the present application to solve the above technical problems is: a method for detecting infrared small targets based on gray scale difference and information entropy, which specifically includes the following steps:

[0007] Step S1, pre-process the original infrared image based on a Scharr gradient operator, and obtain a gradient module feature map and a direction angle feature map of the original infrared image according to a pre-processing result;

[0008] Step S2, construct a direction angle filtering template, filter process the direction angle feature map through the direction angle filtering template, and extract a filtered direction angle feature map of the original infrared image;

[0009] Step S3, construct a double-neighbor block local contrast module, and the double-neighbor block local contrast module includes a center reference area, a middle buffer area and an outer background area;

[0010] The double-neighbor block local contrast module is used to process the gradient module feature map of the original infrared image, and a processed gradient module feature map is obtained;

[0011] Step S4, calculate the entropy weight features of each pixel point in the original infrared image respectively, and obtain an entropy weight feature map of the original infrared image;

[0012] Step S5, integrate the filtered direction angle feature map, the processed gradient module feature map and the entropy weight feature map of the original infrared image, and obtain a final feature map of the original infrared image;

[0013] Step S6, detect small targets in the original infrared image based on the obtained final feature map.

[0014] Further, the specific process of step S1 is as follows:

[0015] Step S11, use a horizontal direction Scharr gradient operator to perform convolution calculation on the original infrared image, and obtain the horizontal direction gradient of each pixel point respectively, and use a vertical direction Scharr gradient operator to perform convolution calculation on the original infrared image, and obtain the vertical direction gradient of each pixel point respectively;

[0016] Step S12, for any one pixel point, calculate the gradient module feature and the direction angle feature of the pixel point according to the horizontal direction gradient and the vertical direction gradient of the pixel point;

[0017] Similarly, after traversing each pixel point in the original infrared image, the gradient module feature map and the direction angle feature map of the original infrared image are obtained.

[0018] Further, the horizontal direction Scharr gradient operator and the vertical direction Scharr gradient operator are respectively as follows:

[0019]

[0020] Wherein, G x_newis a horizontal direction Scharr gradient operator;

[0021]

[0022] wherein G y_new is a vertical direction Scharr gradient operator.

[0023] Further, the gradient module feature and the direction angle feature of the pixel point are calculated in the step S12, and the calculation method is specifically:

[0024] Taking a pixel point (x, y) as an example, the horizontal direction gradient of the pixel point (x, y) is denoted as The vertical direction gradient of the pixel point (x, y) is denoted as The gradient module feature of the pixel point (x, y) is:

[0025]

[0026] wherein M(x, y) is the gradient module feature of the pixel point (x, y);

[0027] The direction angle feature of the pixel point (x, y) is:

[0028]

[0029] wherein D(x, y) is the direction angle feature of the pixel point (x, y).

[0030] Further, the size of the direction angle filter template is 3x3, the element of the first row and the first column in the direction angle filter template is π / 4, the element of the first row and the second column is π / 2, the element of the first row and the third column is 3π / 4, the element of the second row and the first column is 0, the element of the second row and the second column is 0, the element of the second row and the third column is π, the element of the third row and the first column is -π / 4, the element of the third row and the second column is -π / 2, and the element of the third row and the third column is -3π / 4.

[0031] Further, in the step S2, the direction angle feature map is filtered by the direction angle filter template, and a filtered direction angle feature map of the original infrared image is extracted, and the specific process is:

[0032] The direction angle filter template is slid on the direction angle feature map, so that the center of the direction angle filter template sequentially passes through each pixel point on the direction angle feature map.

[0033] For a pixel point (x, y) on the direction angle feature map, a direction angle feature map composed of (x, y) and all pixel points in the 3x3 neighborhood of (x, y) is denoted as D', and the direction angle filter template is used to convolve the direction angle feature map D', and the convolution result is taken as the filtered direction angle feature of the pixel point (x, y):

[0034] GD(x, y) = D' * DFT

[0035] wherein * represents convolution operation, DFT represents the direction angle filter template, and GD(x, y) is the filtered direction angle feature of the pixel point (x, y);

[0036] The filtered direction angle features of all the pixel points form a filtered direction angle feature map of the original infrared image.

[0037] Further, the size of the double-neighborhood block local contrast module is 15x15; wherein:

[0038] The region composed of the pixels located in the first to third rows and the first to third columns in the double-neighborhood block local contrast module is recorded as B o1 ;

[0039] The region composed of the pixels located in the first to third rows and the fourth to twelfth columns in the double-neighborhood block local contrast module is recorded as B o2 ;

[0040] The region composed of the pixels located in the first to third rows and the thirteenth to fifteenth columns in the double-neighborhood block local contrast module is recorded as B o3 ;

[0041] The region composed of the pixels located in the fourth to twelfth rows and the thirteenth to fifteenth columns in the double-neighborhood block local contrast module is recorded as B o4 ;

[0042] The region composed of the pixels located in the thirteenth to fifteenth rows and the thirteenth to fifteenth columns in the double-neighborhood block local contrast module is recorded as B o5 ;

[0043] The region composed of the pixels located in the thirteenth to fifteenth rows and the fourth to twelfth columns in the double-neighborhood block local contrast module is recorded as B o6 ;

[0044] The region composed of the pixels located in the thirteenth to fifteenth rows and the first to third columns in the double-neighborhood block local contrast module is recorded as B o7 ;

[0045] The region composed of the pixels located in the fourth to twelfth rows and the first to third columns in the double-neighborhood block local contrast module is recorded as B o8 ;

[0046] The region composed of the pixels located in the fourth to sixth rows and the fourth to sixth columns in the double-neighborhood block local contrast module is recorded as B i1 ;

[0047] The region composed of the pixels located in the fourth to sixth rows and the seventh to ninth columns in the double-neighbor block local contrast module is denoted as B i2 ;

[0048] The region composed of the pixels located in the fourth to sixth rows and the tenth to twelfth columns in the double-neighbor block local contrast module is denoted as B i3 ;

[0049] The region composed of the pixels located in the seventh to ninth rows and the tenth to twelfth columns in the double-neighbor block local contrast module is denoted as B i4 ;

[0050] The region composed of the pixels located in the tenth to twelfth rows and the tenth to twelfth columns in the double-neighbor block local contrast module is denoted as B i5 ;

[0051] The region composed of the pixels located in the tenth to twelfth rows and the seventh to ninth columns in the double-neighbor block local contrast module is denoted as B i6 ;

[0052] The region composed of the pixels located in the tenth to twelfth rows and the fourth to sixth columns in the double-neighbor block local contrast module is denoted as B i7 ;

[0053] The region composed of the pixels located in the seventh to ninth rows and the fourth to sixth columns in the double-neighbor block local contrast module is denoted as B i8 ;

[0054] The region composed of the pixels located in the seventh to ninth rows and the seventh to ninth columns in the double-neighbor block local contrast module is denoted as T.

[0055] The region B o1 , B o2 , B o3 , B o4 , B o5 , B o6 , B o7 and B o8 composes an outer layer background area, the region B i1 , B i2 , B i3 , B i4 , B i5 , B i6 , B i7 and B i8 composes a middle layer buffer area, and the region T is a center reference area.

[0056] Further, the gradient model feature map of the original infrared image is processed by the double-neighborhood block local contrast module to obtain a processed gradient model feature map; the specific process is as follows:

[0057] The double-neighborhood block local contrast module is slid on the gradient model feature map, so that the center of the double-neighborhood block local contrast module passes through each pixel point on the gradient model feature map in turn, and the processed gradient model feature of each pixel point is obtained, and the processed gradient model feature map is formed by using the processed gradient model features of the pixel points;

[0058] For a pixel point (x, y) on the gradient model feature map, when the center of the double-neighborhood block local contrast module slides to the pixel point (x, y), the calculation method of the processed gradient model feature of the pixel point (x, y) is as follows:

[0059] Step S31, the dissimilarity of each region in the reference region and the middle layer buffer region and the dissimilarity of each region in the reference region and the outer layer background region are calculated.

[0060]

[0061] wherein, d(T, IB i ) represents the dissimilarity of the reference region and the i-th region in the middle layer buffer region; d(T, OB j ) represents the dissimilarity of the reference region and the j-th region in the outer layer background region; m T represents the average of the gradient model features of each pixel point in the reference region; m IBi represents the average of the gradient model features of each pixel point in the i-th region in the middle layer buffer region, i = 1, 2, …, 8; m OBj represents the average of the gradient model features of each pixel point in the j-th region in the outer layer background region, j = 1, 2, …, 8.

[0062] Step S32, the contrast difference of the reference region and the middle layer buffer region and the contrast difference of the reference region and the outer layer background region are calculated according to the result in step S31.

[0063] DTI = min[d(T, IB i ) x d(T, IB i+4 )], i = 1, 2, 3, 4

[0064] DTO = min[d(T, OB j ) x d(T, OB j+4 )], j = 1, 2, 3, 4

[0065] wherein, DTI is the contrast difference of the reference region and the middle layer buffer region, DTO is the contrast difference of the reference region and the outer layer background region, and “x” represents multiplication operation.

[0066] Step S33, calculating the post-processing gradient module feature of the pixel point (x, y) according to the DTI and the DTO:

[0067] DNLMCM = DTI * DTO

[0068] Wherein, DNLMCM is the post-processing gradient module feature of the pixel point (x, y).

[0069] Further, the specific process of the step S4 is:

[0070] Taking the pixel point (i, j) as an example

[0071]

[0072] Wherein, f(i, j) represents the pixel value of the pixel point (i, j); M*N is the local window size of the image with the pixel point (i, j) as the center, p ij is the gray distribution of the pixel point (i, j), H f is the entropy weight feature of the pixel point (i, j);

[0073] The entropy weight features of the global pixel points in the original infrared image form an entropy weight feature map.

[0074] Further, the specific process of the step S5 is:

[0075]

[0076] Wherein, represents pixel-by-pixel multiplication, GD' is the filtered direction angle feature map, DNLMCM' is the post-processing gradient module feature map, H' f is the entropy weight feature map, and FEGSDE is the final feature map.

[0077] The beneficial effects of the present application are:

[0078] The gradient module feature extracted by the present application can make the target and other regions obviously different, the target region can be enhanced by extracting the direction angle feature and combining filtering, background region suppression is realized, the local information entropy is used as a weight factor, which can effectively supplement the information loss caused in the gradient domain feature extraction process. Finally, by integrating the filtered direction angle feature map, the post-processing gradient module feature map and the entropy weight feature map, the accuracy of infrared small target detection can be improved. Since the method of the present application has very good detection effect on small targets in infrared images, the small target detection accuracy and the universality of multiple use environments are improved. BRIEF DESCRIPTION OF DRAWINGS

[0079] Figure 1 is the overall flowchart of the method of the present application;

[0080] Figure 2 schematic diagram of a direction angle filter template;

[0081] Figure 3 schematic diagram of a double-neighbor block local contrast module;

[0082] Figure 4 schematic diagram of a double-neighbor block local contrast module;

[0083] Figure 5 schematic diagram of a real scene applied in the experiment;

[0084] Figure 6 schematic diagram of detection effect under a cloudy background condition;

[0085] Figure 7 schematic diagram of detection effect under a cloudy background condition;

[0086] Figure 8 schematic diagram of detection effect under a ground surface background condition;

[0087] Figure 9 schematic diagram of detection effect under a building background condition. DETAILED DESCRIPTION

[0088] Embodiment I: in combination with Figure 1 This embodiment is described. The infrared small target detection method based on gray scale difference and information entropy described in this embodiment specifically includes the following steps:

[0089] Step S1, pre-processing the original infrared image based on the Scharr gradient operator, and obtaining the gradient module feature map and the direction angle feature map of the original infrared image according to the pre-processing result;

[0090] Step S2, constructing a direction angle filter template, filtering the direction angle feature map through the direction angle filter template, and extracting the filtered direction angle feature map of the original infrared image;

[0091] Step S3, constructing a double-neighbor block local contrast module, and the double-neighbor block local contrast module includes a center reference area, a middle buffer area and an outer background area;

[0092] The double-neighbor block local contrast module is used to process the gradient module feature map of the original infrared image, and a processed gradient module feature map is obtained;

[0093] Step S4, calculating the entropy weight features of each pixel point in the original infrared image respectively, and obtaining the entropy weight feature map of the original infrared image;

[0094] Step S5, the filtered direction angle feature map, the processed gradient module feature map and the entropy weight feature map of the original infrared image are integrated to obtain a final feature map of the original infrared image;

[0095] Step S6, based on the obtained final feature map, small targets in the original infrared image are detected, and the detection process can use existing methods.

[0096] Specific implementation method two: the difference between this embodiment and the specific implementation method one is that the specific process of step S1 is:

[0097] Step S11, the original infrared image is convolved by using a horizontal direction Scharr gradient operator to obtain a horizontal direction gradient of each pixel point, and the original infrared image is convolved by using a vertical direction Scharr gradient operator to obtain a vertical direction gradient of each pixel point;

[0098] Step S12, for any pixel point, the gradient module feature and the direction angle feature of the pixel point are calculated according to the horizontal direction gradient and the vertical direction gradient of the pixel point.

[0099] Similarly, after traversing each pixel point in the original infrared image, a gradient module feature map and a direction angle feature map of the original infrared image are obtained.

[0100] The other steps and parameters are the same as those in the specific implementation method one.

[0101] Specific implementation method three: the difference between this embodiment and the specific implementation method one or two is that the horizontal direction Scharr gradient operator and the vertical direction Scharr gradient operator are respectively:

[0102]

[0103] wherein, G x_new is the horizontal direction Scharr gradient operator;

[0104]

[0105] wherein, G y_new is the vertical direction Scharr gradient operator.

[0106] The other steps and parameters are the same as those in the specific implementation method one or two.

[0107] This embodiment is an improvement on the existing Scharr operator, and the improved Scharr operator can collect more image information when calculating the gradient by increasing the dimension.

[0108] Specific implementation four: the difference between this embodiment and one of the first to third specific implementations is that in the step S12, the calculation method of the gradient module feature and the direction angle feature of the pixel point is specifically as follows:

[0109] Taking a pixel point (x, y) as an example, the horizontal direction gradient of the pixel point (x, y) is denoted as The vertical direction gradient of the pixel point (x, y) is denoted as The gradient module feature of the pixel point (x, y) is:

[0110]

[0111] Wherein, M(x, y) is the gradient module feature of the pixel point (x, y);

[0112] The direction angle feature of the pixel point (x, y) is:

[0113]

[0114] Wherein, D(x, y) is the direction angle feature of the pixel point (x, y).

[0115] The other steps and parameters are the same as one of the first to third specific implementations.

[0116] The target region has a unique gradient feature, that is, the gradient direction angle of the target and its surrounding neighborhood points to the center, and the module of the gradient of the target region is larger; whether it is a complex background or a sharp geometric edge, its gradient feature is obviously different from that of the target region. Therefore, the gradient information of the small target can be used to help locate the edge and contour of the target, so as to better detect and locate.

[0117] Specific implementation five: in combination with Figure 2 This embodiment is described. The difference between this embodiment and one of the first to fourth specific implementations is that the size of the direction angle filtering template is 3x3, the element of the first row and the first column in the direction angle filtering template is π / 4, the element of the first row and the second column is π / 2, the element of the first row and the third column is 3π / 4, the element of the second row and the first column is 0, the element of the second row and the second column is 0, the element of the second row and the third column is π, the element of the third row and the first column is -π / 4, the element of the third row and the second column is -π / 2, and the element of the third row and the third column is -3π / 4.

[0118] The other steps and parameters are the same as one of the first to fourth specific implementations.

[0119] Specific implementation six: the difference between this embodiment and one of the first to fifth specific implementations is that in the step S2, the direction angle feature map is filtered by the direction angle filtering template, and the filtered direction angle feature map of the original infrared image is extracted, and the specific process is as follows:

[0120] The direction angle filtering template is slid on the direction angle feature map, so that the center of the direction angle filtering template sequentially passes each pixel point on the direction angle feature map;

[0121] For a pixel point (x, y) on the direction angle feature map, a direction angle feature map composed of (x, y) and all pixel points in the 3*3 neighborhood of (x, y) is denoted as D', the direction angle filtering template is used to convolve the direction angle feature map D', and the convolution result is taken as the filtered direction angle feature of the pixel point (x, y):

[0122] GD(x, y) = D' * DFT

[0123] Wherein, * represents convolution operation, DFT represents the direction angle filtering template, and GD(x, y) is the filtered direction angle feature of the pixel point (x, y);

[0124] The filtered direction angle features of all pixel points constitute the filtered direction angle feature map of the original infrared image.

[0125] The other steps and parameters are the same as one of the first to fifth embodiments.

[0126] In order to fit the center directivity of the small target block, the angles in the direction angle filtering template are all directed to the center, the direction angle filtering template is used to filter the direction angle feature map, so that the direction angle filtering template has higher correlation with the to-be-processed feature, the value calculated will be larger, and the target area can be enhanced; for the pixel points in other areas, the direction is directed to the direction angle filtering template with low correlation, and the calculated result is relatively low, so that the background area is suppressed.

[0127] It should be noted that when the center of the direction angle filtering template slides to the boundary pixel of the direction angle feature map, the direction angle filtering template will exceed the boundary of the direction angle feature map, at this time, the direction angle feature corresponding to the position exceeding the boundary in the template is considered as 0.

[0128] Specific embodiment seven: combined with Figure 3 and Figure 4 This embodiment is described. This embodiment is different from one of the first to sixth embodiments in that the size of the double-neighborhood block local contrast module is 15*15; wherein:

[0129] The region composed of the pixels located in the first to third rows and the first to third columns in the double-neighborhood block local contrast module is denoted as B o1 ;

[0130] The region composed of the pixels located in the first to third rows and the fourth to twelfth columns in the double-neighborhood block local contrast module is denoted as B o2 ;

[0131] The area consisting of pixels located in the first to third rows and the thirteenth to fifteenth columns in the dual-neighborhood block local contrast module is recorded as B o3 ;

[0132] The area consisting of pixels in the fourth to twelfth rows and the thirteenth to fifteenth columns in the dual-neighborhood block local contrast module is denoted as B. o4 ;

[0133] The area consisting of pixels in the 13th to 15th rows and the 13th to 15th columns in the dual-neighborhood block local contrast module is recorded as B o5 ;

[0134] The area consisting of pixels in the 13th to 15th rows and the 4th to 12th columns in the dual-neighborhood block local contrast module is denoted as B o6 ;

[0135] The area consisting of pixels in the 13th to 15th rows and the 1st to 3rd columns in the dual-neighborhood block local contrast module is recorded as B o7 ;

[0136] The area consisting of pixels in the fourth to twelfth rows and the first to third columns in the dual-neighborhood block local contrast module is recorded as B o8 ;

[0137] The area consisting of pixels located in the fourth to sixth rows and the fourth to sixth columns in the dual-neighborhood block local contrast module is recorded as B i1 ;

[0138] The area consisting of pixels in the fourth to sixth rows and the seventh to ninth columns in the dual-neighborhood block local contrast module is denoted as B i2 ;

[0139] The area consisting of pixels in the fourth to sixth rows and the tenth to twelfth columns in the dual-neighborhood block local contrast module is denoted as B i3 ;

[0140] The area consisting of pixels in the seventh to ninth rows and the tenth to twelfth columns in the dual-neighborhood block local contrast module is denoted as B. i4 ;

[0141] The area consisting of pixels located in the tenth to twelfth rows and the tenth to twelfth columns in the dual-neighborhood block local contrast module is recorded as B i5 ;

[0142] The area consisting of pixels in the tenth to twelfth rows and the seventh to ninth columns in the dual-neighborhood block local contrast module is denoted as B. i6 ;

[0143] The region composed of the pixels located in the seventh to ninth rows and the fourth to sixth columns in the double-neighbor block local contrast module is recorded as B i7 ;

[0144] The region composed of the pixels located in the seventh to ninth rows and the fourth to sixth columns in the double-neighbor block local contrast module is recorded as B i8 ;

[0145] The region composed of the pixels located in the seventh to ninth rows and the fourth to sixth columns in the double-neighbor block local contrast module is recorded as B

[0146] The region B o1 , B o2 , B o3 , B o4 , B o5 , B o6 , B o7 and B o8 composes an outer layer background area, the region B i1 , B i2 , B i3 , B i4 , B i5 , B i6 , B i7 and B i8 composes a middle layer buffer area, and the region T is the central reference area.

[0147] The other steps and parameters are the same as one of the first to sixth embodiments.

[0148] It should be noted that when the center of the module slides to the boundary pixel, i.e., the module exceeds the image boundary, the gradient module feature of the missing pixel is considered to be 0, i.e., 0 is used to participate in the mean value calculation of the gradient module in the region.

[0149] The eighth embodiment is different from one of the first to seventh embodiments in that the double-neighbor block local contrast module is used to process the gradient module feature map of the original infrared image to obtain a processed gradient module feature map, and the specific process is as follows:

[0150] The double-neighbor block local contrast module is slid on the gradient module feature map, so that the center of the double-neighbor block local contrast module sequentially passes through each pixel point on the gradient module feature map, and the processed gradient module feature of each pixel point is obtained, and the processed gradient module features of the pixel points are used to form a processed gradient module feature map;

[0151] For a pixel point (x, y) on the gradient feature map, when the center of the double-neighbor block local contrast module slides to the pixel point (x, y), the calculation method of the processed gradient feature of the pixel point (x, y) is as follows:

[0152] Step S31, calculate the dissimilarity (i.e. the difference in gray scale) of each region in the reference region and the middle layer buffer region and the dissimilarity of each region in the reference region and the outer layer background region.

[0153]

[0154] wherein d(T, IB i ) represents the dissimilarity of the reference region and the i-th region in the middle layer buffer region; d(T, OB j ) represents the dissimilarity of the reference region and the j-th region in the outer layer background region; m T represents the mean value of the gradient feature of each pixel point in the reference region; m IBi represents the mean value of the gradient feature of each pixel point in the i-th region in the middle layer buffer region, i = 1, 2, …, 8; m OBj represents the mean value of the gradient feature of each pixel point in the j-th region in the outer layer background region, j = 1, 2, …, 8.

[0155] Step S32, calculate the contrast difference of the reference region and the middle layer buffer region and the contrast difference of the reference region and the outer layer background region according to the result in step S31.

[0156] DTI = min[d(T, IB i ) x d(T, IB i+4 )], i = 1, 2, 3, 4

[0157] DTO = min[d(T, OB j ) x d(T, OB j+4 )], j = 1, 2, 3, 4

[0158] wherein DTI is the contrast difference of the reference region and the middle layer buffer region, DTO is the contrast difference of the reference region and the outer layer background region, and “x” represents multiplication operation.

[0159] Step S33, calculate the processed gradient feature of the pixel point (x, y) according to DTI and DTO:

[0160] DNLMCM = DTI x DTO

[0161] wherein DNLMCM is the processed gradient feature of the pixel point (x, y).

[0162] The other steps and parameters are the same as one of the first to seventh embodiments.

[0163] In order to make the contrast between the target region and the background region as large as possible, and improve the feature performance of the target region, the present application selects the minimum difference between the reference block and the surrounding background block as the measure of the contrast. The feature saliency of small targets is enhanced by the form of double-layer contrast product, and the gradient modulus contrast information of the center and the neighborhood is effectively utilized, the background area and the strong edge area are effectively suppressed, the feature information of the target area is reserved and enhanced. At the same time, using the gradient modulus mean as the final decision object can effectively reduce the influence of single highlight background or noise.

[0164] Specific embodiment nine: the difference between the present embodiment and one of the specific embodiments one to eight is that the specific process of the step S4 is:

[0165] Taking the pixel point (i, j) as an example

[0166]

[0167] Wherein, f(i,j) represents the pixel value of the pixel point (i, j), the value range is [0, 255]; MxN is the local window size of the image with the pixel point (i, j) as the center, p ij is the gray distribution of the pixel point (i, j), H f is the entropy weight feature of the pixel point (i, j), the base of the logarithm in the present application can be 10;

[0168] The entropy weight features of the global pixel points in the original infrared image form an entropy weight feature map.

[0169] The other steps and parameters are the same as one of the specific embodiments one to eight.

[0170] The present embodiment takes the local information entropy of each pixel point in the original infrared image as a weight factor, by comparing the local information entropy of different regions, it can be determined which regions contain more information and complexity, so as to give them higher weight, and by calculating the entropy of the original infrared image, the information loss caused in the gradient domain feature extraction process can also be effectively supplemented.

[0171] It should be noted that when the center of the local window slides to the boundary pixel, that is, the local window exceeds the image boundary, the pixel value of the missing pixel is considered to be 0, that is, 0 is used to participate in the calculation.

[0172] Specific embodiment ten: the difference between the present embodiment and one of the specific embodiments one to nine is that the specific process of the step S5 is:

[0173]

[0174] Wherein, Indicates pixel-by-pixel multiplication, GD′ is the filtered azimuth feature map, DNLMCM′ is the processed gradient modulus feature map, and H′ f is the entropy weight feature map, and FEGSDE is the final feature map.

[0175] The other steps and parameters are the same as those in Specific Embodiments 1 to 9.

[0176] This implementation integrates the filtered azimuth feature map, the processed gradient modulus feature map, and the entropy weight feature map to form a comprehensive feature representation. This integration method not only considers the local structural information of the image, but also the local statistical characteristics of the image, thereby effectively improving the accuracy of infrared small target feature extraction.

[0177] Experimental part

[0178] The method of the present invention is verified through specific experiments. In order to evaluate the method of the present invention, combined with the application scenario of the small target detection device, different application background images are manually selected in the general infrared small target dataset SIRST, and four environments of cloudy, light cloud, surface, and building are distinguished. The initial images and corresponding label masks in the dataset are reintegrated to establish four sets of infrared sequence experimental image sample sets as the final experimental verification data. Examples of typical samples of each sequence are as follows: Figure 5 As shown in the figure, the target area is highlighted and a local magnification is given. The information of the four sets of sequence images is shown in Table 1:

[0179] Table 1 Sequence image information table

[0180]

[0181] In the comparative experiments, we used seven physical property-based feature extraction baseline methods as comparative references to evaluate the FEGSDE algorithm proposed in this paper. The comparative methods included the TopHat algorithm, MaxMean algorithm, MGDWE algorithm, MPCM algorithm, LIG ​​algorithm, ADMD algorithm, and AADCDD algorithm. Different methods have different parameter settings. The hyperparameter settings of the proposed method and the comparative baseline methods are shown in Table 2.

[0182] Table 2 Algorithm hyperparameter settings

[0183]

[0184] The detection effects of the method of the present invention and other comparison methods under cloudy, lightly cloudy, ground surface and building background conditions are as follows: Figure 6 、 Figure 7 、 Figure 8 and Figure 9 shown.

[0185] To verify the feasibility of the algorithm of the present application, quantitative analysis is performed. In order to more objectively give the detection results of the algorithm, Table 3 gives the quantitative detection evaluation results of eight feature extraction algorithms under four different sequence backgrounds, including background suppression factor BSF value and signal-to-clutter ratio gain SCRG, wherein the bold part is the maximum value.

[0186] Table 3 result comparison table of different algorithms

[0187]

[0188] By comparing the BSF values of the algorithm under different backgrounds, it is found that the BSF obtained under the condition of few clouds is higher, and the BSF under the conditions of cloudy and ground background is slightly lower, which can show that for different small target detection algorithms, the cloud clutter and ground vegetation interference test the background suppression ability of the algorithm more. By selecting a single scene for longitudinal comparison, it is found that the BSF values of the TopHat algorithm, the MaxMean algorithm and the MPCM algorithm based on a single feature are obviously lower than those of the fusion algorithm based on multiple physical features. This can show that from the perspective of multiple features, realizing the complementary advantages between features has a certain influence on improving the algorithm ability.

[0189] By comparing the SCRG values of each algorithm under a certain sequence background, it is found that the SCRG values of the single feature algorithm are lower than those of the multi-feature algorithm, which shows that the enhancement effect of the multi-feature fusion algorithm on the target is better than that of the single feature algorithm. Overall, the FEGSDE algorithm of the present application can obtain better SCRG under different sequence backgrounds, which shows that the algorithm has strong target feature enhancement ability.

[0190] Overall, the FEGSDE algorithm of the present application can obtain obviously better effect in four different sequence backgrounds. Only in the cloudy background, the average BSF value of the FEGSDE algorithm of the present application is not obviously leading compared with other algorithms, while in the other three backgrounds, the BSF value of the FEGSDE algorithm of the present application is obviously large, which can show that the FEGSDE algorithm proposed in the present application has strong suppression ability for different background environments, and can obtain high suppression effect on various background clutter interference.

[0191] The above examples of the present application are only used to illustrate the calculation model and calculation process of the present application, and are not a limitation on the embodiments of the present application. For ordinary skilled persons in the art, other different forms of changes or variations can be made on the basis of the above description, and it is impossible to enumerate all the embodiments here, and any obvious changes or variations derived from the technical solutions of the present application still fall within the protection scope of the present application.

Claims

1. A method for detecting small infrared targets based on grayscale difference and information entropy, characterized in that: The method specifically comprises the following steps: Step S1, preprocessing the original infrared image based on the Scharr gradient operator, and obtaining a gradient modulus feature map and an angular feature map of the original infrared image according to the preprocessing result; Step S2: constructing a directional angle filtering template, filtering the directional angle feature map using the directional angle filtering template, and extracting a filtered directional angle feature map of the original infrared image; Step S3: constructing a dual-neighborhood block local contrast module, wherein the dual-neighborhood block local contrast module includes a central reference area, a middle buffer area, and an outer background area; The dual-neighborhood block local contrast module is used to process the gradient mode feature map of the original infrared image to obtain the processed gradient mode feature map; Step S4, respectively calculating the entropy weight feature of each pixel in the original infrared image to obtain an entropy weight feature map of the original infrared image; Step S5, integrating the filtered azimuth feature map, the processed gradient modulus feature map, and the entropy weight feature map of the original infrared image to obtain a final feature map of the original infrared image; Step S6: Detect small targets in the original infrared image based on the obtained final feature map.

2. The infrared small target detection method based on grayscale difference and information entropy according to claim 1, characterized in that: The specific process of step S1 is: Step S11, performing convolution calculation on the original infrared image using the Scharr gradient operator in the horizontal direction to obtain the horizontal gradient of each pixel point, and performing convolution calculation on the original infrared image using the Scharr gradient operator in the vertical direction to obtain the vertical gradient of each pixel point; Step S12: For any pixel point, calculate the gradient modulus feature and the directional angle feature of the pixel point based on the horizontal gradient and the vertical gradient of the pixel point; Similarly, after traversing each pixel in the original infrared image, the gradient modulus feature map and directional angle feature map of the original infrared image are obtained.

3. The infrared small target detection method based on grayscale difference and information entropy according to claim 2, characterized in that: The Scharr gradient operator in the horizontal direction and the Scharr gradient operator in the vertical direction are respectively: Among them, G x_new is the Scharr gradient operator in the horizontal direction; Among them, G y_new is the Scharr gradient operator in the vertical direction.

4. The infrared small target detection method based on grayscale difference and information entropy according to claim 3 is characterized in that: In step S12, the calculation method of the gradient modulus feature and the directional angle feature of the pixel point is specifically as follows: Taking the pixel point (x, y) as an example, the horizontal gradient of the pixel point (x, y) is recorded as The vertical gradient of the pixel point (x, y) is recorded as Then the gradient modulus feature of the pixel point (x, y) is: Among them, M(x,y) is the gradient modulus feature of the pixel point (x,y); The directional angle feature of the pixel point (x, y) is: Among them, D(x,y) is the direction angle feature of the pixel point (x,y).

5. The infrared small target detection method based on grayscale difference and information entropy according to claim 4 is characterized in that: The size of the directional angle filter template is 3×3, the elements in the first row and first column of the directional angle filter template are π / 4, the elements in the first row and second column are π / 2, the elements in the first row and third column are 3π / 4, the elements in the second row and first column are 0, the elements in the second row and second column are 0, the elements in the second row and third column are π, the elements in the third row and first column are -π / 4, the elements in the third row and second column are -π / 2, and the elements in the third row and third column are -3π / 4.

6. The infrared small target detection method based on grayscale difference and information entropy according to claim 5, characterized in that: In step S2, the directional angle feature map is filtered using the directional angle filtering template to extract the filtered directional angle feature map of the original infrared image. The specific process is as follows: Slide the directional angle filter template on the directional angle feature map so that the center of the directional angle filter template passes through each pixel point on the directional angle feature map in sequence; For the pixel point (x, y) on the directional angle feature map, the directional angle feature map composed of (x, y) and all pixels in the 3×3 neighborhood of (x, y) is recorded as D′. The directional angle filter template is used to convolve the directional angle feature map D′, and the convolution result is used as the filtered directional angle feature of the pixel point (x, y): GD(x,y)=D′*DFT Where * represents convolution operation, DFT represents azimuth filter template, and GD(x,y) is the azimuth feature of pixel (x,y) after filtering. The filtered angular features of all pixels constitute the filtered angular feature map of the original infrared image.

7. The infrared small target detection method based on grayscale difference and information entropy according to claim 6, characterized in that: The size of the dual-neighborhood block local contrast module is 15×15; wherein: The area consisting of pixels located in the first to third rows and the first to third columns in the dual-neighborhood block local contrast module is recorded as B o1 ; The area consisting of pixels in the first to third rows and in the fourth to twelfth columns in the dual-neighborhood block local contrast module is denoted as B o2 ; The area consisting of pixels located in the first to third rows and the thirteenth to fifteenth columns in the dual-neighborhood block local contrast module is recorded as B o3 ; The area consisting of pixels in the fourth to twelfth rows and the thirteenth to fifteenth columns in the dual-neighborhood block local contrast module is denoted as B o4 ; The area consisting of pixels in the 13th to 15th rows and the 13th to 15th columns in the dual-neighborhood block local contrast module is recorded as B o5 ; The area consisting of pixels in the 13th to 15th rows and the 4th to 12th columns in the dual-neighborhood block local contrast module is denoted as B o6 ; The area consisting of pixels in the 13th to 15th rows and the 1st to 3rd columns in the dual-neighborhood block local contrast module is recorded as B o7 ; The area consisting of pixels in the fourth to twelfth rows and the first to third columns in the dual-neighborhood block local contrast module is recorded as B o8 ; The area consisting of pixels located in the fourth to sixth rows and the fourth to sixth columns in the dual-neighborhood block local contrast module is recorded as B i1 ; The area consisting of pixels in the fourth to sixth rows and the seventh to ninth columns in the dual-neighborhood block local contrast module is denoted as B i2 ; The area consisting of pixels in the fourth to sixth rows and the tenth to twelfth columns in the dual-neighborhood block local contrast module is denoted as B i3 ; The area consisting of pixels in the seventh to ninth rows and the tenth to twelfth columns in the dual-neighborhood block local contrast module is denoted as B. i4 ; The area consisting of pixels located in the tenth to twelfth rows and the tenth to twelfth columns in the dual-neighborhood block local contrast module is recorded as B i5 ; The area consisting of pixels in the tenth to twelfth rows and the seventh to ninth columns in the dual-neighborhood block local contrast module is denoted as B. i6 ; The area consisting of pixels in the tenth to twelfth rows and the fourth to sixth columns in the dual-neighborhood block local contrast module is denoted as B i7 ; The area consisting of pixels in the seventh to ninth rows and the fourth to sixth columns in the dual-neighborhood block local contrast module is denoted as B i8 ; The area consisting of pixels located in the seventh to ninth rows and the seventh to ninth columns in the dual-neighborhood block local contrast module is denoted as T; Area B o1 、B o2 、B o3 、B o4 、B o5 、B o6 、B o7 and B o8 Composing the outer background area, area B i1 、B i2 、B i3 、B i4 、B i5 、B i6 、B i7 and B i8 It constitutes the middle buffer zone, and area T is the central reference area.

8. The infrared small target detection method based on grayscale difference and information entropy according to claim 7, characterized in that: The dual-neighborhood block local contrast module is used to process the gradient mode feature map of the original infrared image to obtain the processed gradient mode feature map; the specific process is: Slide the dual-neighborhood block local contrast module on the gradient modulus feature map so that the center of the dual-neighborhood block local contrast module passes through each pixel point on the gradient modulus feature map in sequence, obtain the processed gradient modulus feature of each pixel point, and use the processed gradient modulus features of each pixel point to form a processed gradient modulus feature map; For the pixel point (x, y) on the gradient modulus feature map, when the center of the dual-neighborhood block local contrast module slides to the pixel point (x, y), the calculation method of the processed gradient modulus feature of the pixel point (x, y) is: Step S31, calculating the dissimilarity between the reference area and each area in the middle buffer area, and the dissimilarity between the reference area and each area in the outer background area; Among them, d(T,IB i ) represents the difference between the reference area and the i-th area in the middle buffer zone; d(T,OB j ) represents the difference between the reference area and the jth area in the outer background area; m T Represents the mean value of the gradient modulus characteristics of each pixel in the reference area; m IBi represents the mean value of the gradient modulus feature of each pixel in the i-th region in the middle buffer, i = 1, 2, ..., 8; m OBj represents the mean value of the gradient modulus feature of each pixel in the jth region in the outer background area, j = 1, 2, ..., 8; Step S32, calculating the contrast difference between the reference area and the middle buffer area and the contrast difference between the reference area and the outer background area according to the result in step S31; DTI=min[d(T,IB i )×d(T,IB i+4 )],i=1,2,3,4 DTO=min[d(T,OB j )×d(T,OB j+4 )],j=1,2,3,4 Where DTI is the contrast difference between the reference area and the middle buffer area, DTO is the contrast difference between the reference area and the outer background area, and "×" represents multiplication operation; Step S33: Calculate the processed gradient modulus feature of the pixel point (x, y) based on DTI and DTO: DNLMCM=DTI×DTO Among them, DNLMCM is the processed gradient modulus feature of the pixel point (x, y).

9. The infrared small target detection method based on grayscale difference and information entropy according to claim 8, characterized in that: The specific process of step S4 is as follows: Take pixel (i, j) as an example Among them, f(i,j) represents the pixel value of pixel (i,j); M×N is the size of the local window of the image centered at pixel (i,j), and p ij is the grayscale distribution of pixel (i, j), H f is the entropy weight feature of pixel (i, j); The entropy weight features of global pixels in the original infrared image form an entropy weight feature map.

10. The infrared small target detection method based on grayscale difference and information entropy according to claim 9, characterized in that: The specific process of step S5 is as follows: in, Indicates pixel-by-pixel multiplication, GD′ is the filtered azimuth feature map, DNLMCM′ is the processed gradient modulus feature map, and H′ f is the entropy weight feature map, and FEGSDE is the final feature map.

Images