A method, device, computer equipment and medium for detecting small infrared targets

Abstract

This application belongs to the field of target detection and relates to a method for detecting small infrared targets, comprising: acquiring an infrared image, preprocessing the infrared image to obtain a preprocessed image; performing a multi-directional top-hat transform on the preprocessed image, and simultaneously performing multi-directional weighting to obtain a filtered image; constructing a sliding window structure on the preprocessed image to obtain a preliminary local contrast for each pixel in the preprocessed image; obtaining a final local contrast for each pixel in the filtered image based on the preliminary local contrast of each pixel in the filtered image and the preprocessed image; and obtaining a small infrared target in the infrared image based on the final local contrast of each pixel in the filtered image. This method can improve robustness in irregular, highly undulating, and cluttered backgrounds.

Description

Technical Field

[0001] The present application relates to the field of target detection technology, and in particular to a method, device, computer equipment and medium for detecting small infrared targets. Background Art

[0002] Infrared target detection, as a key part of infrared imaging guidance, is an important technology that determines system performance and is also a very forward-looking content in the field of machine vision.

[0003] Because the characteristics of small infrared targets are less distinct and susceptible to interference from clutter in complex backgrounds, relying solely on a single detection method is insufficient. Currently, detection methods are categorized by the amount of data used, with either single-frame image detection or sequential image detection. Compared to sequential image detection, single-frame image detection imposes fewer restrictions on experimental equipment and places lower demands on the guidance system. Research methods can be categorized into traditional and deep learning approaches.

[0004] Currently, there are many studies on traditional single-frame detection methods. Common traditional methods include: 1) frequency domain filtering method; 2) directional gradient method; 3) method based on low-rank sparse representation, etc.

[0005] However, the above methods cannot cope with complex environmental backgrounds and lack robustness. When the background noise is complex and irregular, or when the infrared target is smaller and has low intensity, this is a major challenge for detection methods under limited data conditions. Summary of the Invention

[0006] Based on this, it is necessary to provide a method, device, computer equipment and medium for detecting infrared weak targets to address the above technical problems, which can improve the robustness in irregular, highly undulating and multi-clutter backgrounds and achieve accurate target detection tasks.

[0007] A method for detecting infrared dim small targets, comprising:

[0008] Acquiring an infrared image, and performing preprocessing on the infrared image to obtain a preprocessed image;

[0009] Performing a multi-directional top-hat transform on the preprocessed image and performing multi-directional weighting to obtain a filtered image;

[0010] constructing a sliding window structure for the preprocessed image to obtain a preliminary local contrast of each pixel in the preprocessed image;

[0011] Obtaining a final local contrast of each pixel in the filtered image based on the preliminary local contrast of each pixel in the filtered image and the preprocessed image;

[0012] A dim target in the infrared image is obtained according to the final local contrast of each pixel in the filtered image.

[0013] In one embodiment, it further includes:

[0014] Calculating a scaling factor based on the weight of each pixel in the preprocessed image and the filtered image;

[0015] According to the pre-processed image, calculating the difference degree, the fluctuation degree and the symmetry degree of each pixel in the pre-processed image, and obtaining a screening factor;

[0016] According to the final local contrast of each pixel in the filtered image, the small target in the infrared image is obtained, which includes:

[0017] Obtaining a dim target in the infrared image according to the scale factor and a final local contrast of each pixel in the filtered image;

[0018] or, obtaining a small target in the infrared image according to the screening factor and a final local contrast of each pixel in the filtered image;

[0019] Alternatively, a small target in the infrared image is obtained according to the scale factor, the screening factor, and a final local contrast of each pixel in the filtered image.

[0020] In one embodiment, the preprocessed image is subjected to a multi-directional top-hat transform and multi-directional weighting to obtain a filtered image, including:

[0021] The fuzzy set theory is used to calculate the membership degree of each pixel in the preprocessed image in the region where it is located, and the weight of each pixel in the preprocessed image is obtained according to the membership degree:

[0022]

[0023]

[0024] Where, f(x,y)| k Indicates the membership of the current pixel (x, y) in the region where the pixel is located when the side length of the central region takes the value k. (x, y) represents the coordinates of the current pixel, k represents the side length of the central region, and I(x, y) represents the value of the current pixel. Represents the average grayscale value of the nine largest pixels in area A centered at (x, y). represents the average value of the area C centered at (x, y), W F (x,y) represents the weight of the current pixel (x,y) in the preprocessed image, and exp represents the exponential function with a natural constant as the base;

[0025] The structural elements of the erosion template are weighted in multiple directions according to the weight of each pixel to obtain the erosion image:

[0026] I e (x,y)=max((IΘe i W F )(x,y)),i=1,...,8

[0027] Where, I e (x,y) represents the value of the current pixel (x,y) after the corrosion operation, Θ represents the corrosion operation, e i Represents the structural element in the i direction;

[0028] Perform the first dilation operation, the second dilation operation, and the difference operation on each pixel of the eroded image to obtain a filtered image:

[0029]

[0030]

[0031]

[0032] Where, I d ′(x,y) represents the max The result of the first dilation operation under the template, I d (x,y) represents the value of the current pixel (x,y) after the second expansion operation. represents the dilation operation, d max Indicates the annular area when L is 8 in the expansion template, d min Indicates the annular area when L is 4 in the expansion template, I t (x,y) represents the filtered image.

[0033] In one embodiment, constructing a sliding window structure for the preprocessed image to obtain a preliminary local contrast of each pixel in the preprocessed image includes:

[0034]

[0035] Where lcm(x,y) represents the preliminary local contrast of the current pixel (x,y) in the preprocessed image, and B min (x,y) represents the minimum average value of the surrounding sub-blocks in the sliding window structure, B max (x,y) represents the maximum average value of the surrounding sub-blocks in the sliding window structure, B i (x, y) represents the average of the three largest pixel grayscale values in the i-th sub-block centered on the current pixel (x, y), where i = 1…8.

[0036] In one embodiment, obtaining the final local contrast of each pixel in the filtered image according to the preliminary local contrast of each pixel in the preprocessed image includes:

[0037]

[0038] Where, LCM T (x,y) represents the final local contrast of the current pixel (x,y) in the filtered image.

[0039] In one embodiment, calculating the scaling factor based on the weight of each pixel in the preprocessed image and the filtered image includes:

[0040] W(x,y)=max{W0,max(W F )·I t (x,y)}

[0041] W0=u0·w(x,y)

[0042]

[0043]

[0044] Where W(x,y) represents the scale factor, W0 represents the initial scale factor, u0 represents the uniformity coefficient of the central sub-block of the sliding window structure, w(x,y) represents the uniformity weight, and u i represents the uniformity coefficient of the i-th sub-block, Represents the center pixel of the i-th sub-block Gray value, mean i Represents the average grayscale value of the i-th sub-block.

[0045] In one embodiment, according to the pre-processed image, calculating the difference degree, the undulation degree, and the symmetry degree of each pixel in the pre-processed image, and obtaining the screening factors includes:

[0046] According to the pre-processed image, the difference degree of each pixel in the pre-processed image is calculated:

[0047] D(x,y)=max{d horizon ,d vertical}

[0048]

[0049]

[0050] Where D(x,y) represents the difference degree within a 3×3 pixel block centered at (x,y), and d horizonIndicates the grayscale value difference between corresponding pixels in the horizontal direction, d vertical Indicates the grayscale value difference between corresponding pixels in the vertical direction;

[0051] According to the pre-processed image, the fluctuation of each pixel in the pre-processed image is calculated:

[0052] s(x,y)=min{mean(T1)-mean(B1),...mean(T4)-mean(B4)}

[0053] Where s(x,y) represents the fluctuation of the area where the pixel (x,y) is located, T i Indicates 1×3 pixels near the center in the corresponding direction, B i Indicates the far center position of 5×3 pixels in the corresponding direction, i=1...4;

[0054] According to the preprocessed image, the symmetry of each pixel in the preprocessed image is calculated:

[0055] d B (x,y)=max{|mean(B1)-mean(B3)|,|mean(B2)-mean(B4)|}

[0056] Where, d B (x,y) represents the symmetry of the grayscale distribution of the background where the pixel (x,y) is located;

[0057] According to the difference degree and symmetry degree, the evaluation coefficient is obtained:

[0058]

[0059] Where E(x,y) represents the evaluation coefficient;

[0060] According to the fluctuation, the gradient coefficient is obtained:

[0061]

[0062] Where S(x,y) represents the gradient coefficient;

[0063] According to the evaluation coefficient and the gradient coefficient, the screening factor is obtained:

[0064]

[0065] Where C(x,y) represents the screening factor.

[0066] A device for detecting small infrared targets, comprising:

[0067] An acquisition module, configured to acquire an infrared image and preprocess the infrared image to obtain a preprocessed image;

[0068] A filtering module, configured to perform a multi-directional top-hat transform on the pre-processed image and simultaneously perform multi-directional weighting to obtain a filtered image;

[0069] A sliding window module, configured to construct a sliding window structure for the preprocessed image to obtain a preliminary local contrast of each pixel in the preprocessed image;

[0070] a contrast module, configured to obtain a final local contrast of each pixel in the filtered image based on the preliminary local contrast of each pixel in the filtered image and the preprocessed image;

[0071] The output module is used to obtain the dim target in the infrared image according to the final local contrast of each pixel in the filtered image.

[0072] A computer device includes a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, the following steps are implemented:

[0073] Acquiring an infrared image, and performing preprocessing on the infrared image to obtain a preprocessed image;

[0074] Performing a multi-directional top-hat transform on the preprocessed image and performing multi-directional weighting to obtain a filtered image;

[0075] constructing a sliding window structure for the preprocessed image to obtain a preliminary local contrast of each pixel in the preprocessed image;

[0076] Obtaining a final local contrast of each pixel in the filtered image based on the preliminary local contrast of each pixel in the filtered image and the preprocessed image;

[0077] A dim target in the infrared image is obtained according to the final local contrast of each pixel in the filtered image.

[0078] A computer-readable storage medium stores a computer program, which, when executed by a processor, implements the following steps:

[0079] Acquiring an infrared image, and performing preprocessing on the infrared image to obtain a preprocessed image;

[0080] Performing a multi-directional top-hat transform on the preprocessed image and performing multi-directional weighting to obtain a filtered image;

[0081] constructing a sliding window structure for the preprocessed image to obtain a preliminary local contrast of each pixel in the preprocessed image;

[0082] Obtaining a final local contrast of each pixel in the filtered image based on the preliminary local contrast of each pixel in the filtered image and the preprocessed image;

[0083] A dim target in the infrared image is obtained according to the final local contrast of each pixel in the filtered image.

[0084] The above-mentioned infrared dim target detection method combines the advantages of morphological filtering and local contrast measurement. Through local feature information analysis, it greatly improves the performance of the single-frame image-based detection method, effectively solves the problem of infrared dim target detection under complex sky background, improves the robustness of the method under irregular, strong undulation, and multi-clutter background, and realizes the task of accurate target detection. BRIEF DESCRIPTION OF THE DRAWINGS

[0085] Figure 1 This is a diagram showing an application scenario of a method for detecting small infrared targets in one embodiment;

[0086] Figure 2 1 is a flow chart of a method for detecting small infrared targets in one embodiment;

[0087] Figure 3 FIG. 1 is a calculation area diagram of degree of membership in one embodiment;

[0088] Figure 4 Figure 2 shows an erosion template and an expansion template in one embodiment, (a) is an erosion template, and (b) is an expansion template;

[0089] Figure 5 A sub-block partition diagram of a sliding window structure in one embodiment;

[0090] Figure 6 Figure 1 is a structural diagram of a screening factor in one embodiment, (a) is a 3×3 pixel block, and (b) is the calculation of local grayscale distribution features;

[0091] Figure 7 Schematic diagram of the architecture of a method for detecting small infrared targets in one embodiment;

[0092] Figure 8 Figure 1 is a test diagram of a method for detecting a small infrared target in one embodiment. (a) is an infrared image, (b) is target detection using a morphological filtering method in the prior art, (c) is target detection using a local contrast measurement method in the prior art, and (d) is target detection using the method of the present application.

[0093] Figure 9Figure 1 is a test diagram of a method for detecting a small infrared target in another embodiment. (a) is an infrared image, (b) is target detection using a morphological filtering method in the prior art, (c) is target detection using a local contrast measurement method in the prior art, and (d) is target detection using the method of the present application.

[0094] Figure 10 Figure 1 is a test diagram of a method for detecting a small infrared target in another embodiment. (a) is an infrared image, (b) is target detection using a morphological filtering method in the prior art, (c) is target detection using a local contrast measurement method in the prior art, and (d) is target detection using the method of the present application.

[0095] Figure 11 1 is a structural block diagram of a device for detecting small infrared targets in one embodiment;

[0096] Figure 12 FIG. 1 is a diagram showing the internal structure of a computer device in one embodiment. DETAILED DESCRIPTION

[0097] In order to make the purpose, technical solutions and advantages of this application more clearly understood, the present application is further described in detail below with reference to the accompanying drawings and examples. It should be understood that the specific embodiments described herein are merely for explaining this application and are not intended to limit this application. All other embodiments obtained by persons of ordinary skill in the art based on the embodiments in this application without creative work are within the scope of protection of this application.

[0098] It should be noted that all directional indications in the embodiments of the present application (such as up, down, left, right, front, back, etc.) are only used to explain the relative position relationship, movement status, etc. between the various components under a certain specific posture (as shown in the accompanying drawings). If the specific posture changes, the directional indication will also change accordingly.

[0099] In addition, the terms "first," "second," and so on, used in this application are for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Therefore, features defined as "first" or "second" may explicitly or implicitly include at least one such feature. In the description of this application, "multiple groups" means at least two groups, such as two groups, three groups, and so on, unless otherwise specifically defined.

[0100] In this application, unless otherwise specified or limited, the terms "connect," "fix," etc. should be understood in a broad sense. For example, "fix" can mean a fixed connection, a detachable connection, or an integral connection; it can mean a mechanical connection, an electrical connection, a physical connection, or a wireless communication connection; it can mean a direct connection or an indirect connection through an intermediate medium; it can mean internal communication between two elements or an interaction between two elements, unless otherwise specified. Those skilled in the art will understand the specific meanings of the above terms in this application based on the specific circumstances.

[0101] In addition, the technical solutions between the various embodiments of the present application can be combined with each other, but it must be based on the fact that ordinary technicians in this field can implement it. When the combination of technical solutions is mutually contradictory or cannot be implemented, it should be deemed that such a combination of technical solutions does not exist and is not within the scope of protection required by this application.

[0102] The present application provides a method for detecting small infrared targets, which can be applied to Figure 1 In the application environment shown, the terminal 102 communicates with the server 104 via a network. The terminal 102 may include but is not limited to various personal computers, laptops, smart phones, tablet computers, and portable wearable devices. The server 104 may be a server corresponding to various portal websites or a backend of a work system.

[0103] This application provides a method for detecting infrared weak targets, such as Figure 2 As shown, in one embodiment, the method is applied to Figure 1 The following example illustrates the terminal in the example, including:

[0104] Step 202: Acquire an infrared image and preprocess the infrared image to obtain a preprocessed image.

[0105] In this step, the infrared image is the original image obtained, and preprocessing refers to performing anisotropic filtering operation on the infrared image to smooth the background, thereby improving the prominence of weak infrared targets. Among them, the anisotropic filtering operation (i.e., diffusion filtering) is an existing technology and will not be repeated here.

[0106] Step 204 , performing a multi-directional top-hat transform on the pre-processed image and performing multi-directional weighting to obtain a filtered image.

[0107] Specifically:

[0108] The fuzzy set theory is used to calculate the membership of each pixel in the preprocessed image in the region where it is located, and the weight of each pixel in the preprocessed image is obtained according to the membership:

[0109]

[0110]

[0111] Where, f(x,y)| k Indicates the membership of the current pixel (x, y) in the region where the pixel is located when the side length of the central region takes the value k. (x, y) represents the coordinates of the current pixel, k represents the side length of the central region, and I(x, y) represents the value of the current pixel. Indicates the side length of the central area where the pixels in area A are centered at (x, y). represents the average value of the area C centered at (x, y), W F (x, y) represents the membership weight of the current pixel (x, y) in the preprocessed image when the k value is taken, and exp represents the exponential function with the natural constant as the base; area A, area C, and k are as follows Figure 3 As shown;

[0112] The structural elements of the erosion template are weighted in multiple directions according to the weight of each pixel, and the top hat transform is improved to obtain the eroded image:

[0113] I e (x,y)=max((IΘe i W F )(x,y)),i=1,...,8

[0114] Where, I e (x,y) represents the value of the current pixel (x,y) after the corrosion operation, Θ represents the corrosion operation, e i Represents the structural element in the i direction;

[0115] Perform the first dilation operation, the second dilation operation, and the difference operation on each pixel of the eroded image to obtain a filtered image:

[0116]

[0117]

[0118]

[0119] Where, I d ′(x,y) represents the max The result of the first dilation operation under the template, I d (x,y) represents the value of the current pixel (x,y) after the second expansion operation. represents the dilation operation, d max Indicates the annular area when L is 8 in the expansion template, d min Indicates the annular area when L is 4 in the expansion template, It (x,y) represents the filtered image.

[0120] In this step, the top hat transform is a typical morphological filter that extracts independent pixel regions by first eroding and then dilating the opening operation. The basic opening operation is expressed as follows:

[0121]

[0122]

[0123] Where I represents the original image, (IΘe) represents the corrosion operation on I, Indicates the expansion operation on I, e and d are both structural elements, m is the offset in the erosion template, n is the offset in the expansion template, I(x,y) represents the grayscale value of the pixel with coordinates (x,y) in the original image, (IΘe)(x,y) represents the value of the pixel with coordinates (x,y) in the original image after erosion, Indicates the value of the pixel with coordinates (x, y) in the original image after expansion (the same below). The erosion template and expansion template established are as follows Figure 4 shown.

[0124] Using fuzzy set theory to calculate the membership and obtain the weight can improve the corrosion effect. The larger the membership, the greater the difference between the pixel and the area it is in, and the less it belongs to the same category, the more likely the target will be detected.

[0125] Multi-directional weighting of the structural elements of the erosion template means performing an erosion operation on each pixel element in different directions in the erosion template and then taking the maximum value, which can effectively deal with edge clutter.

[0126] When improving the top hat transformation, I d (x,y) will be calculated repeatedly until the value remains constant.

[0127] Compared with the traditional filtering method, this step adopts multi-directional weighted morphological filtering, which adds the selection and weighting of corrosion results in different directions. It can more effectively suppress the influence of some small-area background objects and clutter on the results, which is conducive to the extraction of small targets.

[0128] Step 206: construct a sliding window structure for the preprocessed image to obtain a preliminary local contrast of each pixel in the preprocessed image.

[0129] Specifically:

[0130]

[0131] Where lcm(x,y) represents the preliminary local contrast of the current pixel (x,y) in the preprocessed image, and Bmin (x,y) represents the B of the surrounding sub-block in the sliding window structure i (x,y) minimum mean, B max (x,y) represents the B of the surrounding sub-block in the sliding window structure i (x,y) maximum average value, B i (x, y) represents the average of the three largest pixel grayscale values in the i-th sub-block centered on the current pixel (x, y), where i = 1…8.

[0132] In this step, if Figure 5 As shown in FIG, the sliding window structure is divided into 9 sub-blocks, which are shifted from the upper left to the lower right of the image and calculated in sequence, where the sizes of the sub-blocks are selected as 3×3 and 9×9 pixels respectively.

[0133] This contrast calculation fully combines the advantages of the ratio-difference form and can highlight small targets to a greater extent than the general contrast calculation using only one form. In addition, the average of the three largest pixel values in each sub-block is used as the representative value of the block, which can better enhance small targets and suppress the background.

[0134] Step 208 : Obtain a final local contrast of each pixel in the filtered image based on the preliminary local contrast of each pixel in the filtered image and the preprocessed image.

[0135] Specifically:

[0136]

[0137] Where, LCM T (x,y) represents the final local contrast of the current pixel (x,y) in the filtered image.

[0138] In this step, the final filtered image I t (x,y) is used to assist in obtaining the final local contrast LCM T (x,y), which can ensure that small targets are not "annihilated" when they are near the edge of the bright background.

[0139] Step 210 : obtaining the dim targets in the infrared image according to the final local contrast of each pixel in the filtered image.

[0140] Specifically:

[0141] MFLCM(x,y)=LCM T (x,y)

[0142] Where MFLCM(x,y) represents the small target in the infrared image, that is, the value at (x,y).

[0143] In this step, the final local contrast is used as the detection result of the weak target in the infrared image.

[0144] The above-mentioned infrared dim target detection method combines the advantages of morphological filtering and local contrast measurement. Morphological filtering may enhance noise and clutter while enhancing the target, and local contrast measurement may suppress the target while suppressing the background. This method greatly improves the performance of the single-frame image-based detection method through local feature information analysis, effectively solves the problem of infrared dim target detection under complex sky background, improves the robustness of the method in irregular, strongly undulating, and multi-clutter backgrounds, and achieves accurate target detection tasks.

[0145] It should be understood that although Figure 2 The steps in the flowchart are shown in sequence as indicated by the arrows, but these steps are not necessarily executed in the order indicated by the arrows. Unless otherwise specified in this document, there is no strict order restriction for the execution of these steps, and these steps can be executed in other orders. In addition, Figure 2 At least part of the steps may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily executed at the same time, but can be executed at different times. The execution order of these sub-steps or stages is not necessarily sequential, but can be executed in turn or alternately with other steps or at least part of the sub-steps or stages of other steps.

[0146] In one embodiment, it further includes:

[0147] Calculate the scaling factor based on the weight of each pixel in the preprocessed image and the filtered image;

[0148] Calculating the difference, fluctuation, and symmetry of each pixel in the preprocessed image according to the preprocessed image, and obtaining a screening factor;

[0149] According to the final local contrast of each pixel in the filtered image, the weak targets in the infrared image include:

[0150] According to the scale factor and the final local contrast of each pixel in the filtered image, the weak target in the infrared image is obtained;

[0151] Or, based on the screening factor and the final local contrast of each pixel in the filtered image, the weak target in the infrared image is obtained;

[0152] Or, the dim targets in the infrared image are obtained according to the scale factor, the screening factor and the final local contrast of each pixel in the filtered image.

[0153] In this embodiment, the final local contrast can be corrected and improved by using a scaling factor or a screening factor or a combination of the scaling factor and the screening factor to more accurately obtain a small target in the infrared image. That is:

[0154] MFLCM(x,y)=W(x,y)LCM T (x,y)

[0155] or:

[0156] MFLCM(x,y)=C(x,y)LCM T (x,y)

[0157] or:

[0158] MFLCM(x,y)=W(x,y)C(x,y)LCM T (x,y)

[0159] Where W(x,y) represents the scaling factor and C(x,y) represents the screening factor.

[0160] In one embodiment, calculating the scaling factor based on the weight of each pixel in the preprocessed image and the filtered image includes:

[0161] W(x,y)=max{W0,max(W F )·I t (x,y)}

[0162] W0=u0·w(x,y)

[0163]

[0164]

[0165] Where W(x,y) represents the scale factor, W0 represents the initial scale factor, u0 represents the uniformity coefficient of the central sub-block of the sliding window structure, w(x,y) represents the uniformity weight, which can amplify the grayscale feature gap between regular areas and irregular areas, and u i represents the uniformity coefficient of the i-th sub-block, Represents the center pixel of the i-th sub-block Gray value, mean i Represents the average grayscale value of the i-th sub-block.

[0166] In this embodiment, the scaling factor can fully take into account the grayscale distribution characteristics of complex backgrounds and small targets, and adopts sub-block feature difference weighting to ensure that the target can be effectively enhanced and the background can be suppressed.

[0167] In one embodiment, based on the pre-processed image, the difference degree, the undulation degree, and the symmetry degree of each pixel in the pre-processed image are calculated, and the screening factors are obtained, including:

[0168] According to the preprocessed image, calculate the difference degree of each pixel in the preprocessed image:

[0169] D(x,y)=max{d horizon ,d vertical}

[0170]

[0171]

[0172] Where D(x,y) represents the degree of difference (maximum degree of difference) within a 3×3 pixel block centered at (x,y), and d horizon It represents the grayscale value difference between corresponding pixels in the horizontal direction within a 3×3 pixel block centered at (x, y), d vertical Represents the grayscale value difference between corresponding pixels in the vertical direction within a 3×3 pixel block centered at (x, y);

[0173] According to the preprocessed image, the fluctuation degree and directional gradient change rate of each pixel in the preprocessed image are calculated:

[0174] s(x,y)=min{mean(T1)-mean(B1),...mean(T4)-mean(B4)}

[0175]

[0176] Where s(x,y) represents the degree of fluctuation (fluctuation: degree of fluctuation) of the area where the pixel (x,y) is located, and s d (x, y) represents the rate of change of the directional gradient, which is the product of the vertical and horizontal fluctuations of the area where the pixel (x, y) is located. i Indicates the 1×3 pixels near the center position in the i direction, B i Indicates 5×3 pixels at the far center position corresponding to direction i, i = 1...4;

[0177] Based on the preprocessed image, calculate the symmetry of each pixel in the preprocessed image:

[0178] d B (x,y)=max{|mean(B1)-mean(B3)|,|mean(B2)-mean(B4)|}

[0179] Where, d B(x,y) represents the symmetry of the grayscale distribution of the background where the pixel (x,y) is located (symmetry: degree of symmetry);

[0180] According to the difference degree and symmetry degree, the evaluation coefficient is obtained:

[0181]

[0182] Where E(x,y) represents the evaluation coefficient, according to D(x,y) and d B (x,y) is obtained, and the degree of change of the evaluation coefficient can be used to reduce the interference of similar clutter; if d B When (x, y) is less than a certain threshold θ, the pixels around the target center are more symmetrical, and D(x, y) is smaller. When the background is steep, the intensity of weak targets is easily disturbed, resulting in distribution distortion.

[0183] According to the fluctuation, the gradient coefficient is obtained:

[0184]

[0185] Where S(x,y) represents the gradient coefficient, i.e., the edge gradient factor, which is used to distinguish the undulating clutter. The true target satisfies the principle of decreasing gradient from the periphery to the center within a certain local range.

[0186] According to the evaluation coefficient and gradient coefficient, the screening factor is obtained:

[0187]

[0188] Where C(x,y) represents the screening factor.

[0189] Preferably, the screening factor is obtained according to the evaluation coefficient and the gradient coefficient:

[0190]

[0191] The screening factor in this embodiment also takes into account the directional gradient change rate, which can further improve the accuracy of infrared dim small target detection.

[0192] In this embodiment, the clutter similar to the small target may appear in the complex cloud background, and further analysis is made from the perspective of the intensity distribution of the small target. Figure 6 For small targets with an approximate two-dimensional Gaussian distribution, they conform to the isotropic distribution and have a certain symmetrical distribution characteristic in a small local area. Therefore, a screening factor is designed, which is mainly aimed at distinguishing small targets with an approximate two-dimensional Gaussian distribution under a complex background (strong undulating sky and cloud background), which can effectively weaken the interference of complex background clutter on weak targets.

[0193] like Figure 7The method architecture diagram shown in the figure verifies this method, and the specific performance is as follows Figures 8 to 10 As shown in the figure, the box represents the real target position and the circle represents the interference noise. It can be seen that the innovative parts of this method are effective for the detection results: when the environmental background tends to be complex, the advantages are gradually reflected. It is especially suitable for the detection of infrared weak targets that are approximately two-dimensional Gaussian in complex sky environments or ocean environments, such as multi-layer clouds.

[0194] To address the problem that small and weak targets and noise are easily enhanced under strong fluctuating clutter conditions, the present application adopts a morphological filtering processing method based on fuzzy weighting. The morphological filtering result based on fuzzy weighting is used to selectively supplement the result of weighted local contrast measurement to improve the contrast measurement result. The obtained detection result is used as the final detection result. This can effectively amplify the difference, effectively distinguish interference from the target, and can definitely extract the desired object. It is not easy to introduce clutter, ensuring ideal detection results. The contrast method of the present application fully considers the characteristics of local image blocks, calculates local contrast separately by analyzing local directional information to adapt to small targets of different sizes, can achieve good suppression effect on complex backgrounds, and is more suitable for detecting small and weak targets in complex backgrounds. A screening factor is introduced for complex backgrounds (such as the sky), which can further suppress noise and enhance small targets with approximately two-dimensional Gaussian distribution.

[0195] The above-mentioned infrared weak target detection method proposes a method (MFLCM) that combines morphological filtering (MorphologyFilter) and local contrast measurement (Local Contrast Measure) for single-frame image air target detection; the method comprehensively analyzes the local information of infrared weak targets, including grayscale value distribution, gradient characteristics, and background clutter characteristics, and designs and proposes a screening factor. At the same time, fuzziness theory is used as a weighting factor to improve the extraction effect of morphological filtering on small targets. The constructed multi-directional morphological filter is combined with weighted local contrast measurement to improve the algorithm performance, so that the method has better performance under complex and irregular sky backgrounds. This patent has certain reference value for improving the accuracy of infrared imaging guidance and can provide assistance in private industries such as security monitoring and maritime navigation.

[0196] This application also provides a detection device for infrared weak targets, such as Figure 11 As shown, in one embodiment, it includes: an acquisition module 1102, a filtering module 1104, a sliding window module 1106, a comparison module 1108 and an output module 1110, wherein:

[0197] An acquisition module 1102 is configured to acquire an infrared image and preprocess the infrared image to obtain a preprocessed image;

[0198] A filtering module 1104 is configured to perform a multi-directional top-hat transform and multi-directional weighting on the pre-processed image to obtain a filtered image;

[0199] A sliding window module 1106 is configured to construct a sliding window structure for the pre-processed image to obtain a preliminary local contrast of each pixel in the pre-processed image;

[0200] a contrast module 1108 for obtaining a final local contrast of each pixel in the filtered image based on the preliminary local contrast of each pixel in the filtered image and the preprocessed image;

[0201] The output module 1110 is configured to obtain a dim target in the infrared image according to the final local contrast of each pixel in the filtered image.

[0202] In one embodiment, the output module 1110 is further used to: calculate a scaling factor based on the weight of each pixel in the preprocessed image and the filtered image; calculate the difference, undulation and symmetry of each pixel in the preprocessed image based on the preprocessed image, and obtain a screening factor; obtain a small and weak target in the infrared image based on the final local contrast of each pixel in the filtered image, including: obtaining a small and weak target in the infrared image based on the scaling factor and the final local contrast of each pixel in the filtered image; or, obtaining a small and weak target in the infrared image based on the screening factor and the final local contrast of each pixel in the filtered image; or, obtaining a small and weak target in the infrared image based on the scaling factor, the screening factor and the final local contrast of each pixel in the filtered image.

[0203] In one embodiment, the filtering module 1104 is further configured to perform a multi-directional top-hat transform on the pre-processed image and perform multi-directional weighting to obtain a filtered image including:

[0204] The fuzzy set theory is used to calculate the membership degree of each pixel in the preprocessed image in the region where it is located, and the weight of each pixel in the preprocessed image is obtained according to the membership degree:

[0205]

[0206]

[0207] Where, f(x,y)| k Indicates the membership of the current pixel (x, y) in the region where the pixel is located when the side length of the central region takes the value k. (x, y) represents the coordinates of the current pixel, k represents the side length of the central region, and I(x, y) represents the value of the current pixel. Represents the average grayscale value of the nine largest pixels in area A centered at (x, y). represents the average value of the area C centered at (x, y), W F (x,y) represents the weight of the current pixel (x,y) in the preprocessed image, and exp represents the exponential function with a natural constant as the base;

[0208] The structural elements of the erosion template are weighted in multiple directions according to the weight of each pixel to obtain the erosion image:

[0209] I e (x,y)=max((IΘe i W F )(x,y)),i=1,...,8

[0210] Where, I e (x,y) represents the value of the current pixel (x,y) after the corrosion operation, Θ represents the corrosion operation, e i Represents the structural element in the i direction;

[0211] Perform the first dilation operation, the second dilation operation, and the difference operation on each pixel of the eroded image to obtain a filtered image:

[0212]

[0213]

[0214]

[0215] Where, I d ′(x,y) represents the max The result of the first dilation operation under the template, I d (x,y) represents the value of the current pixel (x,y) after the second expansion operation. represents the dilation operation, d max Indicates the annular area when L is 8 in the expansion template, d min Indicates the annular area when L is 4 in the expansion template, I t (x,y) represents the filtered image.

[0216] In one embodiment, the sliding window module 1106 is further configured to construct a sliding window structure for the pre-processed image, and obtain the preliminary local contrast of each pixel in the pre-processed image, including:

[0217]

[0218] Where lcm(x,y) represents the preliminary local contrast of the current pixel (x,y) in the preprocessed image, and Bmin (x,y) represents the minimum average value of the surrounding sub-blocks in the sliding window structure, B max (x,y) represents the maximum average value of the surrounding sub-blocks in the sliding window structure, B i (x, y) represents the average of the three largest pixel grayscale values in the i-th sub-block centered on the current pixel (x, y), where i = 1…8.

[0219] In one embodiment, the contrast module 1108 is further configured to obtain, based on the filtered image and the preliminary local contrast of each pixel in the preprocessed image, a final local contrast of each pixel in the filtered image, including:

[0220]

[0221] Where, LCM T (x,y) represents the final local contrast of the current pixel (x,y) in the filtered image.

[0222] In one embodiment, the output module 1110 is further configured to calculate the scaling factor according to the weight of each pixel in the preprocessed image and the filtered image, including:

[0223] W(x,y)=max{W0,max(W F )·I t (x,y)}

[0224] W0=u0·w(x,y)

[0225]

[0226]

[0227] Where W(x,y) represents the scale factor, W0 represents the initial scale factor, u0 represents the uniformity coefficient of the central sub-block of the sliding window structure, w(x,y) represents the uniformity weight, and u i represents the uniformity coefficient of the i-th sub-block, Represents the center pixel of the i-th sub-block Gray value, mean i Represents the average grayscale value of the i-th sub-block.

[0228] In one embodiment, the output module 1110 is further configured to calculate the difference, undulation, and symmetry of each pixel in the pre-processed image according to the pre-processed image, and obtain the screening factors including:

[0229] According to the pre-processed image, the difference degree of each pixel in the pre-processed image is calculated:

[0230] D(x,y)=max{dhorizon ,d vertical}

[0231]

[0232]

[0233] Where D(x,y) represents the difference degree within a 3×3 pixel block centered at (x,y), and d horizon Indicates the grayscale value difference between corresponding pixels in the horizontal direction, d vertical Indicates the grayscale value difference between corresponding pixels in the vertical direction;

[0234] According to the pre-processed image, the fluctuation of each pixel in the pre-processed image is calculated:

[0235] s(x,y)=min{mean(T1)-mean(B1),...mean(T4)-mean(B4)}

[0236] Where s(x,y) represents the fluctuation of the area where the pixel (x,y) is located, T i Indicates 1×3 pixels near the center in the corresponding direction, B i Indicates the far center position of 5×3 pixels in the corresponding direction, i=1...4;

[0237] According to the preprocessed image, the symmetry of each pixel in the preprocessed image is calculated:

[0238] d B (x,y)=max{|mean(B1)-mean(B3)|,|mean(B2)-mean(B4)|}

[0239] Where, d B (x,y) represents the symmetry of the grayscale distribution of the background where the pixel (x,y) is located;

[0240] According to the difference degree and symmetry degree, the evaluation coefficient is obtained:

[0241]

[0242] Where E(x,y) represents the evaluation coefficient;

[0243] According to the fluctuation, the gradient coefficient is obtained:

[0244]

[0245] Where S(x,y) represents the gradient coefficient;

[0246] According to the evaluation coefficient and the gradient coefficient, the screening factor is obtained:

[0247]

[0248] Where C(x,y) represents the screening factor.

[0249] The specific definition of a device for detecting a small infrared target can be found in the definition of a method for detecting a small infrared target described above and will not be repeated here. Each module in the above-mentioned device can be implemented in whole or in part through software, hardware, or a combination thereof. Each of the above-mentioned modules can be embedded in or independent of a processor in a computer device in the form of hardware, or can be stored in a memory in the form of software in the computer device so that the processor can call and execute the operations corresponding to each of the above modules.

[0250] In one embodiment, a computer device is provided. The computer device may be a terminal, and its internal structure diagram may be as follows: Figure 12 As shown. The computer device includes a processor, a memory, a network interface, a display screen and an input device connected via a system bus. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The network interface of the computer device is used to communicate with an external terminal via a network connection. When the computer program is executed by the processor, a method for detecting weak infrared targets is implemented. The display screen of the computer device can be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer device can be a touch layer covering the display screen, or a button, trackball or touchpad provided on the computer device housing, or an external keyboard, touchpad or mouse.

[0251] Those skilled in the art will understand that Figure 12 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than shown in the figure, or combine certain components, or have a different component arrangement.

[0252] In one embodiment, a computer device is provided, including a memory and a processor. The memory stores a computer program, and the processor implements the steps of the method in the above embodiment when executing the computer program.

[0253] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the steps of the method in the above embodiment are implemented.

[0254] Those skilled in the art will appreciate that all or part of the processes in the above-mentioned embodiments can be implemented by instructing the relevant hardware through a computer program. The computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM).

[0255] The technical features of the above embodiments can be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0256] The above-described embodiments merely represent several implementation methods of the present application. While the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the present invention. It should be noted that a person of ordinary skill in the art may make various modifications and improvements without departing from the spirit of the present invention, all of which fall within the scope of protection of the present invention. Therefore, the scope of protection of the present invention shall be determined by the appended claims.

Claims

1. A method for detecting small infrared targets, characterized in that: include: Acquire an infrared image, and preprocess the infrared image to obtain a preprocessed image; Performing a multi-directional top-hat transform on the preprocessed image and performing multi-directional weighting to obtain a filtered image; Constructing a sliding window structure for the preprocessed image to obtain a preliminary local contrast of each pixel in the preprocessed image; Obtaining a final local contrast of each pixel in the filtered image based on the preliminary local contrast of each pixel in the filtered image and the preprocessed image; Obtaining a small target in the infrared image according to a final local contrast of each pixel in the filtered image; Also includes: Calculating a scaling factor based on the weight of each pixel in the preprocessed image and the filtered image; According to the pre-processed image, calculating the difference degree, the fluctuation degree and the symmetry degree of each pixel in the pre-processed image, and obtaining a screening factor; According to the final local contrast of each pixel in the filtered image, the small target in the infrared image is obtained, which includes: Obtaining a dim target in the infrared image according to the scale factor and a final local contrast of each pixel in the filtered image; or, obtaining a small target in the infrared image according to the screening factor and a final local contrast of each pixel in the filtered image; or, obtaining a dim target in the infrared image according to the scale factor, the screening factor, and a final local contrast of each pixel in the filtered image; According to the pre-processed image, the difference degree, the undulation degree and the symmetry degree of each pixel in the pre-processed image are calculated, and the screening factors are obtained, including: According to the pre-processed image, the difference degree of each pixel in the pre-processed image is calculated: Where, Indicates The difference degree within the 3×3 pixel block centered at Indicates the grayscale value difference between corresponding pixels in the horizontal direction, Indicates the grayscale value difference between corresponding pixels in the vertical direction; According to the pre-processed image, the fluctuation of each pixel in the pre-processed image is calculated: Where, Represents pixels The undulation of the area, express T The area corresponds to 1×3 pixels near the center of the direction. express B The far center position in the corresponding direction of the area is 5×3 pixels, ; According to the preprocessed image, the symmetry of each pixel in the preprocessed image is calculated: Where, Represents pixels The symmetry of the grayscale distribution of the background; According to the difference degree and symmetry degree, the evaluation coefficient is obtained: Where, represents the evaluation coefficient; According to the fluctuation, the gradient coefficient is obtained: Where, represents the gradient coefficient; According to the evaluation coefficient and the gradient coefficient, the screening factor is obtained: Where, Represents the screening factor.

2. The method according to claim 1, characterized in that The pre-processed image is subjected to a multi-directional top-hat transform and multi-directional weighting to obtain a filtered image including: The fuzzy set theory is used to calculate the membership degree of each pixel in the preprocessed image in the region where it is located, and the weight of each pixel in the preprocessed image is obtained according to the membership degree: Where, The side length of the center area where the pixel is located is k The current pixel value The degree of membership in the region, Indicates the coordinates of the current pixel, k represents the side length of the central area, Represents the value of the current pixel, Indicates Centered A The average grayscale value of the 9 largest pixels in the area. Indicates Centered C The average value of the region, Represents the current pixel in the preprocessed image The weight of , exp represents the exponential function with the natural constant as the base; The structural elements of the erosion template are weighted in multiple directions according to the weight of each pixel to obtain the erosion image: Where, Indicates the current pixel The value after the erosion operation is represents the erosion operation, express i Structural elements of direction; Perform the first dilation operation, the second dilation operation, and the difference operation on each pixel of the eroded image to obtain a filtered image: Where, Indicates The result of the first dilation operation under the template, Indicates the current pixel The value after the second expansion operation is represents the dilation operation, Indicates the expansion template L Take the annular area represented by 8 o'clock, Indicates the expansion template L Take the annular area represented by 4 o'clock, represents the filtered image.

3. The method according to claim 1 or 2, characterized in that For the preprocessed image, constructing a sliding window structure to obtain the preliminary local contrast of each pixel in the preprocessed image includes: Where, Represents the current pixel in the preprocessed image The initial local contrast, represents the minimum average value of the surrounding sub-blocks in the sliding window structure, Represents the maximum average value of the surrounding sub-blocks in the sliding window structure, Indicates the current pixel The center i The average of the three largest pixel grayscale values in the sub-block, i =1…8.

4. The method according to claim 3, characterized in that Obtaining a final local contrast of each pixel in the filtered image according to the preliminary local contrast of each pixel in the preprocessed image comprises: Where, Represents the current pixel in the filtered image The final local contrast.

5. The method according to claim 1 or 2, characterized in that Calculating a scaling factor based on the weight of each pixel in the preprocessed image and the filtered image includes: Where, represents the scale factor, represents the initial scale factor, Represents the uniformity coefficient of the central sub-block of the sliding window structure, represents the uniformity weight, Indicates the i The uniformity coefficient of the sub-blocks, Indicates the i The center pixel of the sub-block The gray value of Indicates the i The average grayscale value of the sub-blocks.

6. A device for detecting small infrared targets, characterized in that: The method according to any one of claims 1 to 5, comprising: An acquisition module, configured to acquire an infrared image and preprocess the infrared image to obtain a preprocessed image; A filtering module, configured to perform a multi-directional top-hat transform on the pre-processed image and simultaneously perform multi-directional weighting to obtain a filtered image; A sliding window module, configured to construct a sliding window structure for the preprocessed image to obtain a preliminary local contrast of each pixel in the preprocessed image; a contrast module, configured to obtain a final local contrast of each pixel in the filtered image based on the preliminary local contrast of each pixel in the filtered image and the preprocessed image; The output module is used to obtain the dim target in the infrared image according to the final local contrast of each pixel in the filtered image.

7. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 5 are implemented.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 5 are implemented.

Images