A method and device for detecting small infrared targets based on local binary contrast measurement

Abstract

The present invention discloses a method and device for detecting infrared small targets based on local binary contrast measurement. The method comprises the following steps: (1) designing a new three-layer nested sliding window based on local contrast measurement theory, and simultaneously filling the boundaries of the original thermal infrared image; (2) capturing a local area of ​​the image by the sliding window, and defining the relative number c of image pixels based on the pixel grayscale mean of the central block and the pixel grayscale value of the surrounding neighborhood blocks; (3) defining an eight-bit binary string based on the relative number c and converting it into a decimal local binary contrast, thereby obtaining a local binary contrast measurement image I; (4) utilizing structure tensor theory to design an information filter to suppress the linear sparse structure of image I and enhance small target information; and (5) outputting an infrared target detection result image. The present invention comprehensively utilizes local binary patterns and local contrast measurement theory to effectively suppress background, enhance targets, and improve the detection performance of infrared small targets.

Description

Technical Field

[0001] The present invention relates to the field of image processing, and in particular to a method and device for detecting small infrared targets based on local binary contrast measurement. Background Art

[0002] As we all know, thermal infrared small target detection refers to the detection of small targets in infrared images. Due to the extremely low grayscale, high noise, and cluttered backgrounds of small targets in infrared images, their detection is quite challenging. Currently, a large number of infrared small target detection technologies exist, including traditional detection methods and deep learning-based methods. Most traditional methods face the challenge of balancing the detector's target recognition capabilities and background suppression capabilities. As data-driven methods, deep learning-based methods require large amounts of data, and even with more powerful modules and deeper networks, further improving detection performance is difficult. Therefore, the research of infrared target detection algorithms remains a challenging task.

[0003] Traditional infrared small target detection methods can be categorized as those based on the background consistency assumption, those based on the human visual system, and those based on low-rank sparse decomposition. Methods based on the background consistency assumption assume that the background is highly correlated, while small targets disrupt this local correlation. Consequently, filters are constructed to suppress background and noise, enhancing the target. However, in complex backgrounds and low signal-to-noise ratios, these filters not only enhance the target but also clutter. Low-rank sparse decomposition-based methods assume that the background exhibits low-rank characteristics, as most image regions contain background components with both local and non-local correlations, while the target occupies only a small number of pixels and exhibits sparse characteristics. Methods based on low-quality sparse decomposition include algorithms such as IPI, NRAM, RIPT, LogTFNN, and STT-TRNR. However, most current methods underutilize or even destroy the spatial and temporal information in infrared data, resulting in background reduction or target loss. Methods based on the human visual system aim to design contrast mechanisms that exploit the differences between the target and its surrounding local regions to enhance the target and suppress the background. These methods include detectors such as LCM, MDWCM, GSWLCM, and ELUM. However, these detectors based on the human visual system are highly dependent on the contrast between the target and the background, which can easily cause false alarms when the target is close to the surrounding neighborhood. Summary of the Invention

[0004] To address the shortcomings of the prior art, the present invention aims to provide a method and apparatus for detecting small infrared targets based on local binary contrast metrics, leveraging local contrast metrics and local binary pattern theory. The local binary contrast metric is designed to enhance regions with strong contrast, and an information filter is constructed using structural tensor theory to further suppress background and enhance target saliency, thereby improving the detection performance of small thermal infrared targets and enabling small infrared target detection.

[0005] To achieve the above object, the present invention provides the following technical solutions:

[0006] The present invention discloses a method for detecting small infrared targets based on local binary contrast measurement, which is characterized by comprising the following steps:

[0007] Step 1): Design a sliding window with a three-layer nested structure, where the inner layer is a central block, the middle layer is an isolation ring, and the outermost layer contains eight surrounding neighborhood blocks. At the same time, the original thermal infrared image is border-filled.

[0008] Step 2): Capture the local image block through a sliding window with a three-layer nested structure. Based on the pixel grayscale mean of the central block and the pixel grayscale values ​​of the surrounding neighborhood blocks, define the relative number c of the central pixel point (x, y) of the sliding window;

[0009] Step 3): Define an eight-bit binary string LBS by the relative number c of the pixel point (x, y), and convert the eight-bit binary string LBS into a decimal value to obtain the local binary contrast LBCM of the pixel point (x, y). Move the sliding window pixel by pixel and calculate the local binary contrast LBCM of the central pixel of the sliding window, thereby obtaining the local binary contrast metric image I of the original infrared image;

[0010] Step 4): Using structure tensor theory, an information filter is designed to calculate the weighted harmonic mean H(x,y) of the two eigenvalues ​​of the structure tensor of each pixel (x,y) in the local binary contrast metric image I. The matrix H is obtained from the weighted harmonic mean of each pixel to suppress linear sparse structures and enhance small target information.

[0011] Step 5): Normalize the matrix H to obtain the matrix H′, which is used as the infrared target detection result map T.

[0012] The present invention also discloses an infrared small target detection device based on local binary contrast measurement for implementing the method, which comprises:

[0013] A sliding window structure design module is used to design a sliding window with a three-layer nested structure and perform boundary filling on the original infrared image to capture image blocks of local contrast measurement in the infrared image;

[0014] The local binary contrast measurement operator design module is used to design the local contrast measurement criteria for infrared images and perform local binary contrast measurement using a three-layer nested sliding window and local binary pattern theory.

[0015] The local binary contrast measurement module is used to traverse the infrared image pixel by pixel by sliding the window, perform local binary contrast measurement calculations, and perform infrared small target detection;

[0016] Information filter design module, using structure tensor theory to design information filters, suppress linear sparse structures and enhance targets;

[0017] The target detection result output module is used to output the infrared small target detection result map.

[0018] The beneficial effects of the present invention are:

[0019] 1) This paper fully considers the structural characteristics of small thermal infrared targets and comprehensively considers the energy characteristics of the background and target regions. Based on the theory of local contrast measurement, a new three-layer nested sliding window structure is proposed. The isolation ring effectively reduces the influence of background components on the measured small target features. By adjusting the side lengths of the central block and the eight surrounding neighboring blocks, the accuracy of the algorithm's subsequent local contrast measurement of small targets can be improved.

[0020] 2) This paper designs a novel local binary contrast metric based on the energy difference between the central block and its surrounding neighborhood blocks, drawing on the local contrast metric and the angle difference method of local binary patterns. Furthermore, considering the impact of sparse linear structures on detection, a highly efficient information filter based on structural tensor theory is designed to further enhance the detector's background suppression capabilities, especially for linear structures in the background, and improve its small target enhancement capabilities. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] Figure 1 This is a schematic diagram of the local contrast measurement window structure for small target detection in thermal infrared images according to the present invention;

[0022] Figure 2 Schematic diagram of the structure of the infrared small target detection device of the present invention;

[0023] Figure 3 This is an example frame image of the thermal infrared image sequence used for experimental testing;

[0024] Figure 4 The thermal infrared small target detection results of the example frames of the thermal infrared image sequence;

[0025] Figure 5The detection results of the example frames of the thermal infrared image sequence after being detected by the proposed method, NRAM, ELUM and FAMSIS and the corresponding 3D grayscale visualization images;

[0026] Figure 6 The detection results and corresponding three-dimensional grayscale visualization images of example frames of thermal infrared image sequences after TCNN-NPSTT, ASTTV-NTLA and SRSTT detection. DETAILED DESCRIPTION

[0027] In order to make the purpose, technical solutions and advantages of the present invention clearer, the present invention is described in detail below with reference to specific embodiments. Specific embodiments are described below to simplify the present invention. However, it should be understood that the present invention is not limited to the illustrated embodiments, and that various modifications of the present invention are possible without departing from the underlying principles, and these equivalent forms also fall within the scope defined by the appended claims.

[0028] The following will be combined with the accompanying drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0029] The basic steps of the infrared small target detection method based on local binary contrast measurement in this embodiment of the present invention mainly include:

[0030] Step 1: Based on the local contrast measurement theory, a three-layer nested sliding window is designed. The inner layer is a central block, the middle layer is an isolation ring, and the outermost layer contains eight surrounding neighborhood blocks. At the same time, the original thermal infrared image is border-filled.

[0031] Specifically, based on the structure of the traditional local contrast operator, a new sliding window with three-layer nested structure is designed. Figure 1 As shown, the sliding window contains ten different parts. The inner layer is a central block with a side length of S0, denoted by M0; the middle layer is a block with a width of S isolation The isolation ring is embedded in the sliding window and surrounds the central block; the outermost layer is eight surrounding neighborhood blocks evenly distributed around the isolation ring, represented by M1, M2, ..., M8, and its side length scale is S k , k=1,2,…,8. In this embodiment, the sliding window parameters are designed as follows: the scales of the central block and the surrounding neighborhood blocks are set to S0=S around =3, the width of the isolation ring is S isolation =1, fill scale sThe isolation ring is designed to improve the accuracy of the target contrast measurement in the local area. Since the target edge area or background components are mixed in, the addition of the isolation ring can reduce the impact of the target edge area or other background components on the local contrast measurement process.

[0032] In order to make the pixel points at the edge of the image capable of local contrast measurement, the original thermal infrared image is filled with “0” at the edge, and the filling scale s is

[0033]

[0034] Where [·] represents the floor rounding function, S0 represents the side length scale of the central block M0, S isolation Indicates the width of the isolation ring, S k Indicates the side length scale of the surrounding neighborhood blocks.

[0035] Step 2: Capture the local image block through a sliding window with a three-layer nested structure. Based on the pixel grayscale mean of the central block and the pixel grayscale values ​​of the surrounding neighborhood blocks, define the relative number c of the central pixel point (x, y) of the sliding window;

[0036] Specifically, for the sliding window with three-layer nested structure, the pixel grayscale mean m0 of the central block M0 is

[0037]

[0038] Among them, the side length scale S0=3 of the central block M0, (x, y)∈M0 indicates that the pixel point (x, y) belongs to the central block M0, and f0(x, y) represents the grayscale value of the pixel point (x, y) in the central block M0.

[0039] Based on the pixel grayscale mean of the central block and the surrounding neighborhood blocks, the relative number c of the pixel point (x, y) is defined. Specifically, it is the number of neighborhood blocks whose maximum grayscale value of the surrounding neighborhood blocks is γ times less than the pixel grayscale mean m0 of the central block centered on the pixel point (x, y). The calculation formula of the relative number c is:

[0040] c(x,y)=#{m0>γ·max{f k (i,j)}},(i,j)∈M k ,k=1,2,…,8 (3)

[0041] Among them, #{·} is a counting operator, which is used to obtain the number of neighborhood blocks whose maximum grayscale value of the surrounding blocks is γ times less than the pixel grayscale mean m0 of the central block centered at (x, y), (i, j)∈M k ,k=1,2,…,8 represents the pixel point at the (i,j) position of the kth center block, f k(i, j) represents the grayscale value of the pixel. γ is an adjustable positive constant. In this embodiment, γ is 1.1.

[0042] Step 3: Define an eight-bit binary string LBS based on the relative number c of the pixel point (x, y), and convert the eight-bit binary string LBS into a decimal value to obtain the local binary contrast LBCM of the pixel point (x, y). Move the sliding window pixel by pixel and calculate the local binary contrast LBCM of the central pixel of the sliding window, thereby obtaining the local binary contrast metric image I of the original infrared image;

[0043] Specifically, the eight-bit binary string LBS can be defined by the relative number c of the pixel point (x, y) as

[0044]

[0045] Among them, LBS x,y An eight-bit binary string representing the pixel at position (x,y). c(x,y) represents the relative number of the pixel at position (x,y).

[0046] Then, the eight-bit binary string LBS is converted into a decimal value to obtain the local binary contrast LBCM of the pixel (x, y), which is specifically

[0047]

[0048] The sliding window is moved pixel by pixel and the local binary contrast LBCM of the central pixel of the sliding window is calculated, thereby obtaining the local binary contrast metric image I of the image.

[0049] Step 4: Using structure tensor theory, an information filter is designed to calculate the weighted harmonic mean H(x,y) of the two eigenvalues ​​of the structure tensor of each pixel (x,y) in the local binary contrast metric image I. The matrix H is obtained from the weighted harmonic mean of each pixel to suppress linear sparse structures and enhance small target information.

[0050] Specifically, the structure tensor is used to design an information filter to further suppress the linear sparse structure in the local binary contrast metric image I and enhance the small target information. The structure tensor is calculated as

[0051]

[0052] Among them, * represents the two-dimensional convolution operator, g x,σ and g y,σ Represents the Gaussian kernel function with standard deviation σ=3 in the x-axis direction and y-axis direction respectively. ι Can be calculated as

[0053]

[0054] Among them, g ι represents the Gaussian kernel function with standard deviation ι=3.

[0055] At each pixel (x, y) in the local binary contrast metric image I, the structure tensor J can be calculated ι , which has two eigenvalues ​​(λ1 and λ2), which reveal different image features of pixel points (x, y): 1) flat area, λ1≈λ2≈0; 2) structure edge area, λ1>>λ2≈0; 3) corner area, λ1≥λ2>>0. In order to enhance the point features of the target and suppress the edge structure, the structure tensor J of the image pixel point is ι The weighted harmonic mean of the two eigenvalues ​​is obtained by weighted harmonic mean H0(x,y) of the pixel point (x,y)

[0056]

[0057] Wherein, v1 and v2 are two adjustable positive weights. In this embodiment, v1=1, v2=3.

[0058] In order to prevent the denominator from being zero, the formula (8) is improved to obtain

[0059]

[0060] Where H(x,y) is the improved weighted harmonic mean, and ε is set to 0.001 to prevent the denominator from being zero.

[0061] Step 5: Normalize the image matrix H obtained in step 4) and use the obtained matrix H′ as the infrared target detection result map T to achieve the detection of infrared small targets.

[0062] Specifically, the matrix H obtained in step 4) is normalized to obtain the matrix H′

[0063]

[0064] Among them, H min is the minimum value in the matrix H, max is the maximum value in the matrix H;

[0065] The matrix H′ is used as the infrared small target detection result map T to realize the detection of infrared small targets.

[0066] Corresponding to the aforementioned embodiment of an infrared small target detection method based on local binary contrast measurement, the present invention also provides an embodiment of an infrared small target detection device based on local binary contrast measurement.

[0067] Figure 2 FIG. 1 is a block diagram of an infrared small target detection device based on local binary contrast measurement according to an exemplary embodiment. Figure 2 As shown, the device includes:

[0068] A sliding window structure design module is used to design a sliding window with a three-layer nested structure and perform boundary filling on the original infrared image to capture image blocks of local contrast measurement in the infrared image;

[0069] The local binary contrast measurement operator design module is used to design the local contrast measurement criteria for infrared images and perform local binary contrast measurement using a three-layer nested sliding window and local binary pattern theory.

[0070] The local binary contrast measurement module is used to traverse the infrared image pixel by pixel by sliding the window, perform local binary contrast measurement calculations, and perform infrared small target detection;

[0071] Information filter design module, using structure tensor theory to design information filters, suppress linear sparse structures and enhance targets;

[0072] The target detection result output module is used to output the infrared small target detection result map.

[0073] Regarding the apparatus in the above embodiment, the specific manner in which each module performs operations has been described in detail in the embodiment of the method, and will not be elaborated here.

[0074] As for the device embodiment, since it basically corresponds to the method embodiment, the relevant parts can be referred to the partial description of the method embodiment. The device embodiment described above is only schematic. The various modules in the device are a kind of logical function division. There may be other division methods in actual implementation. For example, multiple modules can be combined or integrated into another unit. Another point is that the connection between the modules shown or discussed can be a communication connection through some interfaces, which can be electrical or other forms. Some or all of the modules can be selected according to actual needs to achieve the purpose of the present application scheme. Ordinary technicians in this field can understand and implement it without paying creative labor. The following uses the disclosed real thermal infrared image sequence as an example to illustrate the specific implementation method to reflect the technical effect of the present invention. The specific steps in the embodiment will not be repeated.

[0075] Example

[0076] The accompanying drawings illustrating the embodiments of the present invention serve to more clearly illustrate the objectives, technical solutions, and advantages of the present invention. It should be noted that the specific embodiments described herein are intended only to illustrate the present invention and are not intended to limit the present invention. Any equivalent substitutions, modifications, and the like made within the methodologies and principles provided by the present invention are intended to be included within the scope of protection of the present invention.

[0077] Next, we used a publicly available infrared image sequence as the research object to conduct a validation of the thermal infrared small target detection method. From a qualitative perspective, the effectiveness of the thermal infrared small target detection method was evaluated by using intuitive small target detection result images and corresponding three-dimensional visual grayscale images. From a quantitative perspective, the performance of the thermal infrared small target detection method was evaluated using the 3D-ROC evaluation index system. Overall, the 3D-ROC evaluation index system can be divided into the target detection capability (TD) index, the background suppression capability (BS) index, and the comprehensive effectiveness evaluation index of the detector. The specific indicators are as follows:

[0078] 1) Target detection capability (TD) indicators include AUC (D,τ) , AUC TD ,in

[0079] AUC (D,τ) Indicates (P D ,τ) curve, which represents the target detection ability of the detector

[0080] 0≤AUC (D,τ) ≤1 (9)

[0081] AUC TD Characterizing the joint detection capabilities of detectors

[0082] 0≤AUC TD =AUC (D,F) +AUC (D,τ) )≤2 (10)

[0083] 2) Background suppression (BS) including AUC (F,τ) , AUC BS , AUC SNPR ,in

[0084] AUC (F,τ) Indicates (P D ,τ) curve, which represents the background suppression ability of the detector.

[0085] 0≤AUC (F,τ) ≤1 (11)

[0086] AUC BS Characterizing the joint background suppression capability of detectors

[0087] -1≤AUC BS =AUC (D,F) -AUC (F,τ) )≤1 (12)

[0088] AUC SNPR Characterizing the detector's signal-to-noise ratio

[0089]

[0090] 3) The comprehensive effectiveness indicators of the detector include AUC TDBS , AUC (D,F) , AUC ODP ,in

[0091] AUC TDBS Characterizes the detector's comprehensive target detection and background suppression capabilities

[0092] -1≤AUC TDBS =AUC (D,τ) -AUC (F,τ) )≤1 (14)

[0093] AUC (D,F) Characterization (P D ,P F ) curve, indicating the effectiveness of the detector

[0094] 0≤AUC (D,F) ≤1 (15)

[0095] AUC ODP Characterizes the total detection probability of the detector

[0096] -1≤AUC ODP =AUC (D,F) +AUC (D,τ) -AUC (F,τ) ≤2 (16)

[0097] For a publicly available, real thermal infrared image sequence containing 100 frames, the sequence features: a slow-moving aircraft in a complex mountain-land background, with bright buildings, a low contrast between the target and the surrounding background, suspected targets, and linear edge structures. The frame image size is 256×256. Figure 3 is an example frame of an infrared image sequence, and the corresponding small target detection result obtained by the method of the present invention is shown in FIG. Figure 4As shown in the figure, the small target detection results intuitively show that the complex background containing bright buildings and strong noise is greatly suppressed, and the small target is well enhanced and highlighted. Table 1 shows the quantitative indicators of the small target detection results using NRAM, ELUM, FAMSIS, TCNN-NPSTT, ASTTV-NTLA, SRSTT, and the proposed method for an example sequence of thermal infrared images. The bold and underlined values ​​represent the corresponding optimal AUC results and suboptimal AUC results, respectively.

[0098] The sources of the comparison methods are as follows:

[0099] NRAM comes from L.Zhang, L.Peng, T.Zhang, S.Cao, and Z.Peng, “Infrared smalltarget detection via non-convex rank approximation minimization jointl 2, 1norm,”Remote Sens.,vol.10,no.11,p.1821,2018.

[0100] ELUM comes from E. Zhao, W. Zheng, M. Li, H. Sun, and J. Wang, "Infrared small target detection using local component uncertainty measure with consistency assessment," IEEE Geosci.Remote Sens. Lett., vol.19, pp.1–5, 2022.

[0101] FAMSIS comes from Y.Chen, G.Zhang, Y.Ma, JUKang, and C.Kwan, “Small infraredtarget detection based on fast adaptive masking and scaling with iterativesegmentation,” IEEE Geosci.Remote Sens.Lett., vol.19, pp.1–5, 2021.

[0102] TCNN-NPSTT comes from G.Wang, B.Tao,

[0103] ASTTV-NTLA comes from T.Liu, J.Yang, B.Li, C.Xiao, Y.Sun, Y.Wang, and W.An, "Nonconvex tensor low-rank approximation for infrared small target detection," IEEE Trans.Geosci.Remote Sens., vol.60, pp.1-18, 2021.

[0104] SRSTT comes from J.Li, P.Zhang, L.Zhang, and Z.Zhang, "Sparse regularization-based spatial-temporal twist tensor model for infrared small target detection," IEEE Trans.Geosci.Remote Sens., 2023.

[0105] Table 1 Quantitative indicators of detection results of thermal infrared image example sequences using NRAM, ELUM, FAMSIS, TCNN-NPSTT, ASTTV-NTLA, SRSTT and the proposed method

[0106] method <![CDATA[AUC (D，F) ]]> <![CDATA[AUC (D，τ) ]]> <![CDATA[AUC (F，τ) ]]> <![CDATA[AUC TD ]]> <![CDATA[AUC BS ]]> <![CDATA[AUC sNPR ]]> <![CDATA[AUC TDBS ]]> <![CDATA[AUC oDP ]]> The present invention 1.000 0.987 5.261e-5 1.987 1.000 1.876e4 0.987 1.987 NRAM 1.000 <![CDATA[ 0.996 ]]> 2.954e-4 <![CDATA[ 1.996 ]]> 1.000 3.372e3 <![CDATA[ 0.996 ]]> <![CDATA[ 1.996 ]]> ELUM 1.000 0.932 2.694e-4 1.932 1.000 3.458e3 0.931 1.931 FAMSIS 1.000 0.998 1.152e-4 1.998 1.000 8.660e3 0.998 1.998 TCNN-NPSTT 1.000 0.956 <![CDATA[ 7.039e-5 ]]> 1.956 1.000 <![CDATA[ 1.358e4 ]]> 0.956 1.956 ASTTV-NTLA 1.000 0.947 1.375e-1 1.947 <![CDATA[ 0.863 ]]> 6.887 0.809 1.809 SRSTT 1.000 0.977 4.323e-4 1.977 1.000 2.261e3 0.977 1.977

[0107] Figure 5 A comparison of small target detection results and corresponding 3D grayscale visualization images for a sample sequence of thermal infrared images using the present invention, NRAM, ELUM, and FAMSIS; Figure 6Comparisons of small target detection results and corresponding 3D grayscale visualizations for a sample sequence of thermal infrared images using TCNN-NPSTT, ASTTV-NTLA, and SRSTT are shown. Qualitative results show that while SRSTT can detect targets, the target intensity is somewhat attenuated, resulting in weak target detection capabilities and residual clutter in the detection results. ASTTV-NTLA has very weak background suppression capabilities and residual building components, resulting in a high false alarm rate. While NRAM, ELUM, TCNN-NPSTT, and FAMSIS can detect small targets, their background suppression capabilities are weak. The proposed method, however, can completely suppress background components and enhance the saliency of small targets. The quantitative results of the eight 3D-ROC evaluation metrics shown in Table 1 demonstrate that the proposed method outperforms all other compared methods in terms of background suppression capabilities. Although the target detection-related metrics are lower than those of NRAM and FAMSIS, it still outperforms the other compared methods. Based on the comprehensive target detection capability (TD) metric, background suppression capability (BS) metric, and detector effectiveness evaluation metrics, the proposed method generally outperforms existing methods. From the above qualitative and quantitative analysis, it can be seen that the thermal infrared small target detection method proposed in the present invention has excellent target detection capability, background suppression capability and comprehensive effectiveness.

[0108] The accompanying drawings illustrating the embodiments of the present invention serve to more clearly illustrate the objectives, technical solutions, and advantages of the present invention. It should be noted that the specific embodiments described herein are intended only to illustrate the present invention and are not intended to limit the present invention. Any equivalent substitutions, modifications, and the like made within the methodologies and principles provided by the present invention are intended to be included within the scope of protection of the present invention.

Claims

1. A method for detecting small infrared targets based on local binary contrast measurement, characterized in that: The steps include: Step 1): Design a sliding window with a three-layer nested structure, where the inner layer is a central block, the middle layer is an isolation ring, and the outermost layer contains eight surrounding neighborhood blocks. At the same time, the original thermal infrared image is border-filled. Step 2): Capture the local image block through a sliding window with a three-layer nested structure. Based on the pixel grayscale mean of the central block and the pixel grayscale values ​​of the surrounding neighborhood blocks, define the relative number c of the central pixel point (x, y) of the sliding window; Step 3): Define an eight-bit binary string LBS by the relative number c of the pixel point (x, y), and convert the eight-bit binary string LBS into a decimal value to obtain the local binary contrast LBCM of the pixel point (x, y). Move the sliding window pixel by pixel and calculate the local binary contrast LBCM of the central pixel of the sliding window, thereby obtaining the local binary contrast metric image I of the original infrared image; The step 3) is specifically as follows: defining an eight-bit binary string LBS by the relative number c of the pixel point (x, y) to characterize the local module features, specifically: Among them, LBS x,y An eight-bit binary string representing the pixel at position (x,y), where c(x,y) represents the relative number of the pixel at position (x,y); Then, the eight-bit binary string LBS is converted into a decimal value to obtain the local binary contrast LBCM of the pixel (x, y), which is specifically LBCM(x,y)=2 c -1 (5) Move the sliding window pixel by pixel and calculate the local binary contrast LBCM of the central pixel of the sliding window, thereby obtaining the local binary contrast metric image I of the image; Step 4): Using structure tensor theory, an information filter is designed to calculate the weighted harmonic mean H(x,y) of the two eigenvalues ​​of the structure tensor of each pixel (x,y) in the local binary contrast metric image I. The matrix H is obtained from the weighted harmonic mean of each pixel to suppress linear sparse structures and enhance small target information. Step 5): Normalize the matrix H to obtain the matrix H′, which is used as the infrared target detection result map T.

2. The infrared small target detection method based on local binary contrast measurement according to claim 1 is characterized in that: The step 1) is specifically as follows: A sliding window with a three-layer nested structure is designed. The inner layer is a central block with a side length of S0, represented by M0; the middle layer is a block with a width of S isolation The isolation ring is embedded in the sliding window and surrounds the central block; the outermost layer is eight surrounding neighborhood blocks evenly distributed around the isolation ring, represented by M1, M2, ..., M8, and its side length scale is S k ,k=1,2,…,8; In order to make the local contrast measurement of the pixels at the edge of the image possible, the original thermal infrared image is filled with "0" at the edge, and the filling scale s is Where [·] represents the floor rounding function, S0 represents the side length scale of the central block M0, S isolation Indicates the width of the isolation ring, S k Indicates the side length scale of the surrounding neighborhood blocks.

3. The infrared small target detection method based on local binary contrast measurement according to claim 1 is characterized in that: The step 2) is specifically as follows: Based on the pixel grayscale mean m0 of the central block M0 and the pixel grayscale values ​​of the surrounding neighborhood blocks, the relative number c of the pixel point (x, y) is defined. The calculation formula of the relative number c is: c(x,y)=#{m0>γ·max{f k (i,j)}},(i,j)∈M k ,k=1,2,…,8 (3) Among them, #{·} is a counting operator, which is used to obtain the number of neighboring blocks whose maximum grayscale value of the surrounding neighboring blocks is γ times less than the pixel grayscale mean m0 of the central block M0 centered at (x, y), (i, j)∈M k ,k=1,2,…,8 represents the pixel point at the (i,j) position of the kth center block, f k (i, j) represents the grayscale value of the pixel, and γ is an adjustable positive constant.

4. The infrared small target detection method based on local binary contrast measurement according to claim 1 is characterized in that: The step 4) is specifically as follows: Using the structural tensor theory, the information filter is designed and the structural tensor is calculated as Among them, * represents the two-dimensional convolution operator, g x,σ and g y,σ Represents the Gaussian kernel function with standard deviation σ in the x-axis direction and y-axis direction respectively; thus the structure tensor J ι Calculated as Among them, g ι represents the Gaussian kernel function with standard deviation ι; At each pixel (x, y) in the local binary contrast metric image I, the structure tensor J can be calculated ι , which has two eigenvalues ​​λ1 and λ2. For the flat area, λ1≈λ2≈0; for the structure edge area, λ1>>λ2≈0; for the corner area, λ1≥λ2>>0; The structure tensor J for each pixel (x, y) of the image ι The weighted harmonic mean of the two eigenvalues ​​is obtained by weighted harmonic mean H0(x,y) of the pixel point (x,y) as follows: Among them, v1 and v2 are two adjustable positive weights; In order to prevent the denominator from being zero, the formula (8) is improved to obtain Where H(x,y) is the improved weighted harmonic mean, and ε is a small constant to prevent the denominator from being zero; The matrix H is obtained by the improved weighted harmonic mean H(x,y) of each pixel in the image I, which can suppress the linear sparse structure and enhance the small target information.

5. The infrared small target detection method based on local binary contrast measurement according to claim 1 is characterized in that: The step 5) is specifically as follows: Normalize the matrix H obtained in step 4) to get the matrix H′ Among them, H min is the minimum value in the matrix H, max is the maximum value in the matrix H; The matrix H′ is used as the infrared small target detection result map T to realize the detection of infrared small targets.

6. An infrared small target detection device based on local binary contrast measurement implementing the method of claim 1, characterized in that: include: A sliding window structure design module is used to design a sliding window with a three-layer nested structure and perform boundary filling on the original infrared image to capture image blocks of local contrast measurement in the infrared image; The local binary contrast measurement operator design module is used to design the local contrast measurement criteria for infrared images and perform local binary contrast measurement using a three-layer nested sliding window and local binary pattern theory. The local binary contrast measurement module is used to traverse the infrared image pixel by pixel by sliding the window, perform local binary contrast measurement calculations, and perform infrared small target detection; Information filter design module, using structure tensor theory to design information filters, suppress linear sparse structures and enhance targets; The target detection result output module is used to output the infrared small target detection result map.

Images