Infrared dim small target detection method based on local standard deviation-contrast joint
Patent Information
- Application Number
- CN202611488584.0
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-09-23
- Publication Date
- 2026-10-27
AI Technical Summary
[0004]有鉴于此，本发明创造旨在提供一种基于局部标准差-对比度联合的红外弱小目标检测方法，以解决现有算法无法提取目标尺度特征的不足，不能够适应不同尺度目标的检测需求，本发明将局部标准差作为目标尺度的主要度量，将局部对比度作为目标强度的主要度量，在提高检测精度、增加检测效率的同时，能够有效克服现有算法无法提取目标尺度特征的不足，并能够适应不同尺度目标的检测需求
本发明创造所述的基于局部标准差-对比度联合的红外弱小目标检测方法，可完整保留红外弱小目标尺度信息，适配多尺度目标检测场景，兼顾检测精度与运算效率，计算资源开销低，具备良好的工程实践应用价值。
Smart Images
Figure CN122989751A_ABST
Abstract
Description
Technical Field
[0001] This invention belongs to the field of infrared image target detection technology, and particularly relates to an infrared weak target detection method based on local standard deviation-contrast combination. Background Technology
[0002] Infrared imaging detection boasts advantages such as strong penetration, low vulnerability to revealing its own location, and resistance to enemy interference. In modern applications, infrared search and track detection technology has become a crucial component in precision navigation, early warning, and remote sensing operations. However, as the detection distance increases, the detected target often appears as a weak, speckled feature in infrared images, making it highly susceptible to various background interferences, thus increasing the difficulty of target detection. Therefore, researching algorithms capable of effectively detecting weak infrared targets in complex environments is of great significance for both practical engineering applications and academic research.
[0003] In the field of infrared weak target detection, numerous algorithms have emerged, which can be broadly categorized into three types: spatial filtering, rank-based decomposition, and deep learning methods. Spatial filtering algorithms primarily enhance the visibility of weak infrared targets by increasing their response intensity while suppressing the background, thereby achieving target detection. However, these algorithms cannot account for the true shape of the target and are easily affected by background interference similar to the target's local features. Rank-based decomposition algorithms typically transform the detection problem into a low-rank decomposition problem to distinguish the target from the background. These algorithms are prone to suboptimal solutions, severe target shrinkage effects, and incomplete preservation of the target's scale. Deep learning algorithms heavily rely on computational resources and a large number of samples. These methods demand high-performance computing platforms, but in practical engineering applications, available computing power is often limited, leading to decreased algorithm performance and operating efficiency. Summary of the Invention
[0004] In view of this, the present invention aims to provide an infrared weak target detection method based on local standard deviation-contrast combination, in order to solve the shortcomings of existing algorithms that cannot extract target scale features and cannot adapt to the detection needs of targets of different scales. The present invention uses local standard deviation as the main measure of target scale and local contrast as the main measure of target intensity. While improving detection accuracy and increasing detection efficiency, it can effectively overcome the shortcomings of existing algorithms that cannot extract target scale features and can adapt to the detection needs of targets of different scales.
[0005] To achieve the above objectives, the technical solution created by this invention is implemented as follows: A method for detecting weak infrared targets based on local standard deviation-contrast joint method includes the following steps: S1: Calculate the local standard deviation of the infrared image to obtain the local standard deviation image. Fill the holes in the local standard deviation image to obtain the local standard deviation filled image. Use a Gaussian difference filter to filter the local standard deviation filled image to obtain the local standard deviation processed image containing target scale information. S2: The infrared image is filtered using a local contrast operator template to obtain the target region image and the background region image. The target neighborhood image is calculated based on the background region image. The local contrast-processed image containing target intensity information is obtained by calculating the contrast difference between the target neighborhood image and the target region image. S3: The image processed by local standard deviation and the image processed by local contrast are fused to obtain a confidence image. The detection of small infrared targets is achieved based on the gray values of the confidence image. S4: Based on the detection results obtained in step S3, perform image segmentation on the input infrared image with the weak infrared target as the foreground and other content as the background.
[0006] Furthermore, step S1 specifically includes: S11: Set a local sliding window and slide the local sliding window pixel by pixel on the infrared image. Calculate the gray standard deviation of the pixels in the area covered by the local sliding window at each position. Based on the gray standard deviation of the local sliding window at all positions, obtain the local standard deviation image. S12: Traverse the local standard deviation image, fill the holes in the local standard deviation image, and obtain the local standard deviation filled image; S13: Apply Gaussian difference filtering to the locally standard deviation-filled image to obtain a locally standard deviation-processed image containing target scale information. ; in, Image processing for local standard deviation, For convolution operations, The standard deviation of the Gaussian kernel. Fill the image with local standard deviations. and It is a two-dimensional Gaussian function, where k is the scale coefficient. These are pixel coordinates.
[0007] Furthermore, in step S11, the size of the local sliding window is set to 3×3.
[0008] Furthermore, in step S12, the pixel mean of the local standard deviation image is set to a, 1.2a is used as the high threshold, 0.5a is used as the low threshold, pixels with gray values higher than 1.2a are used as bright pixels, pixels with gray values lower than 0.5a are used as low pixels, and dark pixels surrounded by bright pixels are used as holes.
[0009] Furthermore, the specific steps of step S2 are as follows: S21: Set the local contrast operator template, which includes... There are 1 sub-regions, each with a size of 1. The labels of each sub-region adopt a column-wise repetitive arrangement, with the central sub-region as the target region and all other sub-regions as background regions. S22: Set the coefficients of the target region to the mean filter coefficients, and set the coefficients of the background region to 0 to construct a target extraction template. Use the target extraction template to filter the infrared image to obtain the target region image. ; S23: Set the coefficients of the background region labeled i to the mean filter coefficients, and set the coefficients of other sub-regions to 0 to construct the i-th background extraction template. Use the i-th background extraction template to filter the infrared image to obtain the i-th background region image. ; S24: Replace the background region labeled i with the background region labeled i+1, and repeat step S23 until the background region image corresponding to each background region is obtained. S25: Select four sets of background regions on the four corners of the local contrast operator template. Each set of background regions is arranged in a 2×2 pattern. For each set of background regions, take the maximum pixel value at the same pixel position to generate four first-class target neighborhood images. Four sets of background regions are selected from the four sides of the local contrast operator template. Each set of background regions is arranged in a T-shape and does not include the background regions located at the four corners of the local contrast operator template. For the background region image corresponding to each set of background regions, the maximum pixel value at the same pixel position is taken to generate four second-class target neighborhood images. S26: For four first-class target neighborhood images and four second-class target neighborhood images, extract the pixels at the same pixel position in each target neighborhood image, select the four pixels with the largest pixel value and calculate the average value to generate a target neighborhood mean image. S27: Calculate local contrast in the image using the following formula: ; in, To process the image for local contrast, For the target region image, For the target neighborhood mean image, For pixel coordinate index.
[0010] Furthermore, in step S3, the image processed by local standard deviation and the image processed by local contrast are fused using the following formula: ; in, For confidence level images, Image processing for local standard deviation, Image processing for local contrast.
[0011] Furthermore, in step S3, the maximum grayscale value of the confidence image is set to b, and the pixel corresponding to the grayscale value of 0.5b is taken as the infrared weak target.
[0012] Compared with the prior art, the present invention can achieve the following beneficial effects: The infrared weak target detection method based on local standard deviation-contrast combination described in this invention can completely preserve the scale information of infrared weak targets, adapt to multi-scale target detection scenarios, balance detection accuracy and computational efficiency, have low computational resource overhead, and have good engineering application value. Attached Figure Description
[0013] The accompanying drawings, which form part of this invention, are used to provide a further understanding of the invention. The illustrative embodiments and descriptions of the invention are used to explain the invention and do not constitute an undue limitation of the invention. In the drawings: Figure 1 A schematic flowchart of the infrared weak target detection method based on local standard deviation-contrast joint method described in the embodiments of the present invention; Figure 2 The infrared image described in the embodiments of the present invention; Figure 3 A three-dimensional schematic diagram of the infrared image described in the embodiments of the present invention; Figure 4 The local standard deviation image described in the embodiments of the present invention; Figure 5 The local standard deviation filled image described in the embodiments of the present invention; Figure 6 The local standard deviation processed image is created according to the embodiments of the present invention; Figure 7 A three-dimensional schematic diagram of the local standard deviation processed image as described in the embodiments of the present invention; Figure 8 A schematic diagram of the structure of the local contrast operator template described in the embodiment of the present invention; Figure 9 The local contrast processing image described in the embodiments of the present invention; Figure 10 A three-dimensional schematic diagram of the local contrast processing image described in the embodiments of the present invention; Figure 11The confidence image described in the embodiments of the present invention; Figure 12 A three-dimensional schematic diagram of the confidence image described in the embodiment of the present invention. Detailed Implementation
[0014] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and do not constitute a limitation thereof.
[0015] It should be noted that, unless otherwise specified, the embodiments and features described in the present invention can be combined with each other.
[0016] In the description of this invention, it should be understood that the terms "center," "longitudinal," "lateral," "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "top," "bottom," "inner," and "outer," etc., indicating orientations or positional relationships based on the orientations or positional relationships shown in the accompanying drawings, are only for the convenience of describing this invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation on this invention. Furthermore, the terms "first," "second," etc., are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of indicated technical features. Thus, features defined with "first," "second," etc., may explicitly or implicitly include one or more of that feature. In the description of this invention, unless otherwise stated, "a plurality of" means two or more.
[0017] In the description of this invention, it should be noted that, unless otherwise explicitly specified and limited, the terms "installation," "connection," and "linking" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal connection of two components. Those skilled in the art will understand the specific meaning of the above terms in this invention based on the specific circumstances.
[0018] The present invention will now be described in detail with reference to the accompanying drawings and embodiments.
[0019] like Figure 1 As shown, this invention proposes an infrared weak target detection method based on local standard deviation-contrast joint method, which specifically includes the following steps: S1: Calculate the local standard deviation of the infrared image to obtain the local standard deviation image. Fill the holes in the local standard deviation image to obtain the local standard deviation filled image. Use a Gaussian difference filter to filter the local standard deviation filled image to obtain the local standard deviation processed image containing target scale information. S2: The infrared image is filtered using a local contrast operator template to obtain the target region image and the background region image. The target neighborhood image is calculated based on the background region image. The local contrast-processed image containing target intensity information is obtained by calculating the contrast difference between the target neighborhood image and the target region image. S3: The image processed by local standard deviation and the image processed by local contrast are fused to obtain a confidence image. The detection of small infrared targets is achieved based on the gray values of the confidence image. S4: Based on the detection results obtained in step S3, perform image segmentation on the input infrared image with the weak infrared target as the foreground and other content as the background.
[0020] It should be noted that: 1) Obtaining the local standard deviation processed image based on the local standard deviation algorithm: First, calculate the local standard deviation of the input infrared image to obtain the local standard deviation image; then, find holes in the local standard deviation image, i.e., dark pixel regions surrounded by brighter pixels, and fill the holes; finally, use a Gaussian difference filter to filter the filled image to obtain the local standard deviation processed image, which contains the target's scale information. 2) Obtaining the local contrast processed image based on the local contrast algorithm: First, filter the input infrared image within a local region using a local contrast operator template to obtain the target region image and the background region image respectively; calculate the target neighborhood image from the specified background region image; then, calculate the contrast difference between the target region image and the target neighborhood mean image to obtain the local contrast processed image, which contains the target's intensity information. 3) Obtain the final calculation result: Fuse the two images processed based on local standard deviation and the image processed based on local contrast to obtain a confidence map based on the combination of local standard deviation and local contrast. Then, perform adaptive threshold segmentation on the confidence map to obtain the target detection result.
[0021] Example 1 I. Obtaining images processed using the local standard deviation algorithm.
[0022] 1) Calculate the local standard deviation of the input infrared image. like Figures 2-3 As shown, for the input infrared image Take a sample from the infrared image. A local sliding window is defined, and the standard deviation of pixel grayscale values within the local sliding window is calculated to characterize the pixel grayscale fluctuations and texture complexity within the local sliding window, resulting in a local standard deviation image. In the local standard deviation image In the image, the value of each pixel is the input infrared image. The surrounding area of the corresponding pixel The standard deviation of pixels within the neighborhood.
[0023] The formula for obtaining the local standard deviation image is: ; in, for The mean within the neighborhood is expressed as: ; in, This is the row offset. This is the column offset.
[0024] 2) Fill holes in the local standard deviation image. like Figures 4-5 As shown, in the local standard deviation image In the process, we find dark pixel hole regions surrounded by brighter pixels and fill the holes to obtain a filled local standard deviation image. : ; Where H is the set of holes.
[0025] 3) such as Figures 6-7 As shown, this is an image filled with the local standard deviation after hole filling. The formula for calculating the difference of Gaussian filtering is as follows: ; in, Represents convolution operation. The standard deviation of the Gaussian kernel. and It is a two-dimensional Gaussian function, where k is the scale coefficient. This is the output image of the Gaussian difference filter, i.e., the image processed by local standard deviation; the scale of the difference Gaussian filter is... The standard deviation of the Gaussian kernel is taken as: The scale ratio factor is taken as: Scale parameters .
[0026] II. Obtaining images processed based on local contrast algorithms.
[0027] 1) Obtain the target region image and the background region image. The local contrast operator template for the infrared weak target extraction algorithm based on local contrast is as follows: Figure 8 As shown, with Arranged in a certain way, each sub-region is the size of Therefore, the scale of the local contrast operator template is The size of the sub-region is _____. Therefore, the scale of the local contrast operator template is The central area is T, representing the target area, and the area surrounding T represents the background area B.
[0028] In the 25 sub-regions of the local contrast operator template, only the coefficient of the target region T is a mean filter coefficient, while the coefficients of the other sub-regions are 0. In this case, the target extraction template is used to process the input infrared image. Filtering is performed to obtain the target region image. .
[0029] Similarly, when only the coefficient of the background region B1 is the mean filter coefficient, and the coefficients of other sub-regions are 0, the set first background extraction template is used to process the input infrared image. Filtering is performed to obtain the first background region image. .
[0030] When only the coefficient of the background region B2 is a mean filter coefficient, and the coefficients of other sub-regions are 0, the second background extraction template is used to process the input infrared image. Filtering is performed to obtain the image of the second background region. .
[0031] By analogy, we can obtain... ~ , ~ .
[0032] 2) Calculate the target neighborhood image from the specified background region image. Background area image , , , The maximum value among four pixels at the same position is taken as the neighborhood of the first type of target image. The pixel value is represented as: ; Background area image , , , The maximum value among four pixels at the same position is taken as the neighborhood image of the second type of target. The pixel value is represented as: ; Background area image , , , The maximum value among four pixels at the same position is taken as the neighborhood of the first type of target image. The pixel value is represented as: ; Background area image , , , The maximum value among four pixels at the same position is taken as the neighborhood image of the second type of target. The pixel value is represented as: ; Background area image , , , The maximum value among four pixels at the same position is taken as the neighborhood of the first type of target image. The pixel value is represented as: ; Background area image , , , The maximum value among four pixels at the same position is taken as the neighborhood image of the second type of target. The pixel value is represented as: ; Background area image , , , The maximum value among four pixels at the same position is taken as the neighborhood of the first type of target image. The pixel value is represented as: ; Background area image , , , The maximum value among four pixels at the same position is taken as the neighborhood image of the second type of target. The pixel value is represented as: ; Acquire images based on local contrast processing For two types of target neighborhood images , ... The average of the four largest pixel values at the same pixel position is expressed as: ; in, express ~ The image containing the first i largest pixel values at the same position. This represents the average of the first four images with the highest pixel values.
[0033] like Figures 9-10 As shown, the output image processed based on the local contrast algorithm can be represented as: ; III. Obtain infrared weak target detection results based on the combination of local standard deviation and local contrast.
[0034] like Figures 11-12 As shown, the above and The images are fused to obtain the final output confidence image. , is represented as: ; Among them, operators This represents the dot product operation.
[0035] Set the maximum grayscale value of the confidence image to b, and use the pixel corresponding to the grayscale value of 0.5b as the infrared weak target.
[0036] Fourth, based on the detection results obtained in step S3, the input infrared image is segmented with the weak infrared target as the foreground and other content as the background.
[0037] It should be understood that the various forms of processes shown above can be used to reorder, add, or delete steps. For example, the steps described in this invention disclosure can be executed in parallel, sequentially, or in different orders, as long as the desired result of the technical solution disclosed in this invention can be achieved, and this is not limited herein.
[0038] The specific embodiments described above do not constitute a limitation on the scope of protection of this invention. Those skilled in the art should understand that various modifications, combinations, sub-combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this invention should be included within the scope of protection of this invention.
Claims
1. A method for detecting weak infrared targets based on a combination of local standard deviation and contrast, characterized in that: Specifically, the steps include the following: S1: Calculate the local standard deviation of the infrared image to obtain the local standard deviation image. Fill the holes in the local standard deviation image to obtain the local standard deviation filled image. Use a Gaussian difference filter to filter the local standard deviation filled image to obtain the local standard deviation processed image containing target scale information. S2: Filter the infrared image using a local contrast operator template to obtain the target region image and the background region image. Calculate the target neighborhood image based on the background region image. And obtain a locally contrast-processed image containing target intensity information by calculating the contrast difference between the target neighborhood image and the target region image. S3: The image processed by local standard deviation and the image processed by local contrast are fused to obtain a confidence image. The detection of small infrared targets is achieved based on the gray values of the confidence image. S4: Based on the detection results obtained in step S3, perform image segmentation on the input infrared image with the weak infrared target as the foreground and other content as the background.
2. The infrared weak target detection method based on local standard deviation-contrast joint as described in claim 1, characterized in that: Step S1 specifically includes: S11: Set a local sliding window and slide the local sliding window pixel by pixel on the infrared image. Calculate the gray standard deviation of the pixels in the area covered by the local sliding window at each position. Based on the gray standard deviation of the local sliding window at all positions, obtain the local standard deviation image. S12: Traverse the local standard deviation image, fill the holes in the local standard deviation image, and obtain the local standard deviation filled image; S13: Apply Gaussian difference filtering to the locally standard deviation-filled image to obtain a locally standard deviation-processed image containing target scale information. ； in, Image processing for local standard deviation, For convolution operations, The standard deviation of the Gaussian kernel. Fill the image with local standard deviations. and It is a two-dimensional Gaussian function, where k is the scale coefficient. These are pixel coordinates.
3. The infrared weak target detection method based on local standard deviation-contrast combination according to claim 2, characterized in that: In step S11, the size of the local sliding window is set to 3×3.
4. The infrared weak target detection method based on local standard deviation-contrast joint method according to claim 2, characterized in that: In step S12, the pixel mean of the local standard deviation image is set to a, 1.2a is used as the high threshold, 0.5a is used as the low threshold, pixels with gray values higher than 1.2a are used as bright pixels, pixels with gray values lower than 0.5a are used as low pixels, and dark pixels surrounded by bright pixels are used as holes.
5. The infrared weak target detection method based on local standard deviation-contrast joint as described in claim 1, characterized in that: The specific steps of step S2 are as follows: S21: Set the local contrast operator template, which includes... There are 1 sub-regions, each with a size of 1. The labels of each sub-region adopt a column-wise repetitive arrangement, with the central sub-region as the target region and all other sub-regions as background regions. S22: Set the coefficients of the target region to the mean filter coefficients, and set the coefficients of the background region to 0 to construct a target extraction template. Use the target extraction template to filter the infrared image to obtain the target region image. ; S23: Set the coefficients of the background region labeled i to the mean filter coefficients, and set the coefficients of other sub-regions to 0 to construct the i-th background extraction template. Use the i-th background extraction template to filter the infrared image to obtain the i-th background region image. ; S24: Replace the background region labeled i with the background region labeled i+1, and repeat step S23 until the background region image corresponding to each background region is obtained. S25: Select four sets of background regions on the four corners of the local contrast operator template. Each set of background regions is arranged in a 2×2 pattern. For each set of background regions, take the maximum pixel value at the same pixel position to generate four first-class target neighborhood images. Four sets of background regions are selected from the four sides of the local contrast operator template. Each set of background regions is arranged in a T-shape and does not include the background regions located at the four corners of the local contrast operator template. For the background region image corresponding to each set of background regions, the maximum pixel value at the same pixel position is taken to generate four second-class target neighborhood images. S26: For four first-class target neighborhood images and four second-class target neighborhood images, extract the pixels at the same pixel position in each target neighborhood image, select the four pixels with the largest pixel value and calculate the average value to generate a target neighborhood mean image. S27: Calculate local contrast in the image using the following formula: ； in, To process the image for local contrast, For the target region image, For the target neighborhood mean image, For pixel coordinate index.
6. The infrared weak target detection method based on local standard deviation-contrast joint as described in claim 1, characterized in that: In step S3, the image processed by local standard deviation and the image processed by local contrast are fused using the following formula: ； in, For confidence level images, Image processing for local standard deviation, Image processing for local contrast.
7. The infrared weak target detection method based on local standard deviation-contrast joint as described in claim 1, characterized in that: In step S3, the maximum grayscale value of the confidence image is set to b, and the pixel corresponding to the grayscale value of 0.5b is taken as the infrared weak target.