A method for infrared small target detection based on saliency and weighted guided filtering

Abstract

The present invention discloses a method for detecting infrared small targets based on saliency and weighted guided filtering. The method comprises the following steps: first, obtaining a prior image of the infrared small target using a saliency filtering method; then, performing weighted guided filtering on the original image using four different sets of parameters to obtain an edge-preserving smoothed sequence image; selecting corresponding pixels from the edge-preserving smoothed sequence image and the original image according to the grayscale of the prior image as a background estimation image; performing a subtraction between the original image and the background estimation image to obtain a target image, and outputting a target detection result using a threshold segmentation method. The present invention proposes, for the first time, a method for background estimation and detection of infrared small targets that combines saliency and weighted guided filtering. The method has a simple implementation process and good real-time performance. It can effectively improve the probability of detecting small targets in complex scenes and reduce the probability of false alarms caused by edges, etc.

Description

Technical Field

[0001] The present invention belongs to the field of infrared image processing and target detection, and in particular relates to an infrared small target detection method based on saliency and weighted guided filtering. Background Art

[0002] Infrared detection features passive detection, strong penetration, and 24 / 7 operation. It is widely used in military applications such as early warning and search, infrared guidance, and coastal and air defense, as well as civilian applications such as outdoor rescue, wildlife monitoring, and bird surveillance at airports. As a key component of infrared detection systems, infrared small target detection technology can locate and detect distant targets through image analysis and processing. However, small infrared targets have characteristics such as small imaging area, no distinct textures, no fixed form, and low signal-to-noise ratio. Furthermore, background clutter such as trees, mountains, clouds, and ocean waves often interferes, making accurate detection of small infrared targets difficult.

[0003] Currently, single-frame-based infrared small target detection methods are widely favored by researchers due to their relative simplicity, low hardware requirements, and ease of engineering. Common single-frame detection methods include background estimation, human visual system methods, and low-rank sparse decomposition methods. Background estimation methods typically use spatial or frequency domain filters to suppress the background, which is computationally inefficient and simple to implement, but is easily affected by complex background interference. The human visual system primarily relies on local contrast mechanisms, which are effective when the target meets the visual saliency assumption, but clutter with significant characteristics can cause false alarms. Low-rank sparse decomposition methods exploit the non-local correlation of the background and the sparsity of the target, transforming the target detection problem into an optimization problem. They are effective for detecting small, weak targets with low signal-to-noise ratios, but suffer from high computational complexity and difficulty in engineering. Moreover, most algorithms lack the ability to resist interference from edges.

[0004] In summary, the current infrared small target detection algorithm has shortcomings to varying degrees, such as low robustness, poor real-time performance, and high false alarm rate. There is still a need to seek an efficient and practical infrared small target detection algorithm. Summary of the Invention

[0005] The purpose of the present invention is to overcome the shortcomings of the existing technology and provide an infrared small target detection method based on saliency and weighted guided filtering. The method can effectively suppress false alarms caused by complex background, especially edge clutter, and has the characteristics of low algorithm complexity and strong real-time performance, which can easily meet the high efficiency and real-time requirements in engineering applications.

[0006] The technical solution of the present invention is: a method for detecting small infrared targets based on saliency and weighted guided filtering, the method comprising the following steps:

[0007] Step 1.1: Obtain the original infrared image D and use the saliency filtering method to calculate the infrared small target prior image C;

[0008] Step 1.2: Use four different sets of filtering parameters to perform weighted guided filtering on the original infrared image D to obtain four edge-preserving smoothed sequence images;

[0009] Step 1.3: According to the grayscale distribution of the prior image C, select the corresponding pixels from the edge-preserving smoothed sequence image and the original infrared image D as the background estimation image B;

[0010] Step 1.4: Subtract the original infrared image D from the background estimation image B to obtain the target image T. Use the threshold segmentation method to determine the position of the small infrared target and output the target detection result.

[0011] Furthermore, in the saliency filtering method described in step 1.1, the filter is a 5×5 spatial filter kernel;

[0012] Furthermore, the filtering parameters described in step 1.2 are the window radius and the regularization parameter;

[0013] Furthermore, in the weighted guided filtering described in step 1.2, the weight of a single pixel is the average ratio of the variance of the pixel to the variance of each pixel in the image, and the weight w k The calculation formula is:

[0014]

[0015] Where N is the number of pixels in the image, Var is the variance of the pixels, and ε is a constant;

[0016] Furthermore, in the threshold segmentation described in step 1.4, the product of the maximum value of the target image T and the constant coefficient m is used as the segmentation threshold, and the value range of m is 0.3 to 0.5.

[0017] The advantages of the present invention compared with the prior art are:

[0018] (1) Using visual saliency as the prior information for background estimation greatly reduces the difficulty of background estimation, improves the accuracy of background estimation, and is conducive to improving the detection rate of targets;

[0019] (2) Due to the edge-preserving smoothing characteristics of guided filtering and the addition of edge-aware weights, the use of weighted guided filtering to achieve accurate background estimation will result in less background residue in the target image, especially with good anti-interference ability against edge clutter;

[0020] (3) Both the saliency filtering method and the weighted guided filtering method have the characteristics of simple implementation and good parallelism, and are particularly suitable for applications with high real-time requirements. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] Figure 1 Schematic diagram of the process of infrared small target detection method based on saliency and weighted guided filtering of the present invention;

[0022] Figure 2 This is an infrared image containing a small target exemplified by the present invention;

[0023] Figure 3 The present invention is Figure 2 Constructed target saliency prior map;

[0024] Figure 4 The present invention is Figure 2 Edge-preserving smoothed sequence graph obtained by weighted guided filtering;

[0025] Figure 5 The present invention is Figure 2 、 Figure 3 、 Figure 4 The calculated background image and target image;

[0026] Figure 6 The present invention is Figure 5 The detection result of the target image in is obtained by threshold segmentation. DETAILED DESCRIPTION

[0027] To make the objectives, technical solutions, and advantages of the present invention more clearly understood, the present invention is further described in detail below with reference to the accompanying drawings and examples. The present invention, for the first time, combines saliency filtering with weighted guided filtering to effectively estimate the background image, reduce the impact of clutter such as edges on the detection results, and achieve rapid and accurate detection of small infrared targets.

[0028] The following is a detailed description of an infrared small target detection method based on saliency and weighted guided filtering:

[0029] 1. Get the original infrared image D, such as Figure 2 shown.

[0030] 2. Calculate the saliency prior map C of the infrared small target, such as Figure 3 The specific steps include:

[0031] 2.1. Use a 2×2 mean filter to smooth the image;

[0032] 2.2. Apply 5×5 spatial filter kernel F to the whole image after mean filtering to obtain the confidence map R. The spatial filter kernel F is derived based on the facet model (see A fast-saliency method for real-time infrared small target detection. Infrared Physics & Technology [J]. 2016, 77: 440-450) and is in the following form:

[0033]

[0034] 2.3. Perform grayscale square operation on each pixel in R to obtain the enhanced confidence map E;

[0035] 2.4. Perform grayscale morphological dilation on image E to obtain the prior image C. Grayscale morphological dilation can be found in Digital Image Processing (Rafael C. Gonzalez and Richard E. Woods, Electronic Industry Press, 3rd edition). The structure element SE used is a 5×5 disk in the following form:

[0036]

[0037] 3. Using the weighted guided filtering method (see Weighted Guided Image Filtering. IEEE Transactions on Image Processing [J]. 2015, 24 (1): 120-129), the edge-preserving smoothing sequence graph is calculated under four different parameter conditions, such as Figure 4 The specific steps include:

[0038] 3.1. Calculate the edge perception weight of each pixel, weight w k is the average ratio of the variance of the pixel to the variance of each pixel in the image, and the calculation formula is:

[0039]

[0040] Where N is the number of pixels in the image, Var is the local variance in an 11×11 neighborhood centered on the pixel, and ε is a constant related to the dynamic range L of the original image. Its calculation formula is as follows:

[0041] ε=(0.001×L) 2

[0042] 3.2. Take the window radius r and calculate the mean μ of each pixel in the (2r+1)×(2r+1) neighborhood kand variance And get the linear coefficient a k and b k , the calculation formula is as follows:

[0043]

[0044] b k =(1-a k )μ k

[0045] Among them, λ is the regularization parameter, w k is the edge-aware weight described in 3.1.

[0046] 3.3. Linear coefficient a for each pixel position k and b k , take the window radius r, average in the neighborhood of (2r+1)×(2r+1), and get the average value and Then use the gray value D of the corresponding position of the original image i , get the filtered image gray value q i , the calculation formula is as follows:

[0047]

[0048]

[0049]

[0050] Among them, \r| is the number of pixels within the window radius r, and the size is (2r+1)×(2r+1).

[0051] 3.4, take four different sets of window radius and regularization parameters (r j ,λ j )(j=1,2,3,4), repeat steps 3.1 to 3.3 to obtain the filtering results q under four different parameter conditions j (j=1,2,3,4).

[0052] 4. According to the grayscale distribution of the prior graph C, the edge-preserving smoothing sequence graph q j (j=1,2,3,4) and select the corresponding pixels in the original image D as the background estimation map B, such as Figure 4 The specific steps include:

[0053] 4.1. Normalize the prior image C so that the grayscale distribution range is [0, 1].

[0054] 4.2. According to the gray value of each pixel in the prior image C and the set threshold t j(j=1,2,3,4), segmentally estimate the grayscale value of the background image. The specific calculation method is as follows:

[0055]

[0056] 5. Subtract the original infrared image D from the background estimation image B to obtain the target image T. The normalized target image T is as follows: Figure 5 shown.

[0057] 6. Perform threshold segmentation on the target image T, determine the position of the infrared small target, and output the target detection result, such as Figure 6 As shown in Figure 1, the segmentation threshold is the product of the maximum value of the target image T and the constant coefficient m, and the value range of m is 0.3 to 0.5.

[0058] The above description is merely a preferred embodiment of the present invention and is not intended to limit the scope of protection of the present invention.

Claims

1. A method for detecting small infrared targets based on saliency and weighted guided filtering, characterized in that: The method comprises the following steps: Step 1.1: Obtain the original infrared image D and use the saliency filtering method to calculate the infrared small target prior image C; Step 1.2: Use four different sets of filtering parameters to perform weighted guided filtering on the original image D to obtain four edge-preserving smoothed sequence images; Step 1.3: According to the grayscale distribution of the prior image C, select the corresponding pixels from the edge-preserving smoothed sequence image and the original infrared image D as the background estimation image B; Step 1.4: Subtract the original infrared image D from the background estimation image B to obtain the target image T. Use the threshold segmentation method to determine the position of the small infrared target and output the target detection result.

2. The infrared small target detection method based on saliency and weighted guided filtering according to claim 1, characterized in that: The saliency filtering method described in step 1.1: the filter is a 5×5 spatial filter kernel.

3. The infrared small target detection method based on saliency and weighted guided filtering according to claim 1, characterized in that: The filtering parameters described in step 1.2 are: window radius and regularization parameter.

4. The infrared small target detection method based on saliency and weighted guided filtering according to claim 1, characterized in that: Weighted guided filtering as described in step 1.2: The weight of a single pixel is the average ratio of the variance of the pixel to the variance of each pixel in the image, and the weight w k The calculation formula is: Where N is the number of pixels in the image, Var is the variance of the pixels, and ε is a constant.

5. The infrared small target detection method based on saliency and weighted guided filtering according to claim 1, characterized in that: Threshold segmentation described in step 1.4: The product of the maximum value of the target image T and the constant coefficient m is used as the segmentation threshold, and the value range of m is 0.3 to 0.5.

Images