Infrared small target detection method combining local gravitation and horizontal line connectivity

Abstract

The invention discloses an infrared small target detection method combining local gravitation and horizontal line connectivity, and belongs to the technical field of infrared small target detection, and the method comprises the steps: S1, obtaining an original infrared image # imgabs0 #, and calculating local gravitation # imgabs1 # and # imgabs2 # of pixels in a neighborhood acting on a middle pixel; s2, calculating a multi-direction local gravity contrast according to the local gravity obtained in the step S1; s3, calculating a horizontal line connectivity contrast according to the local gravity obtained in the step S1; s4, performing regional fusion on the multi-direction local gravitation contrast obtained in the step S2 and the horizontal line connectivity contrast obtained in the step S3 to obtain a fusion contrast; and S5, according to the fusion contrast obtained in the step S4, performing segmentation by adopting an adaptive threshold method to obtain a final detection result. According to the technical scheme, background suppression can be improved while the target intensity is kept, and the detection precision is remarkably improved.

Description

Technical Field

[0001] The present invention relates to the technical field of infrared small target detection, and in particular to an infrared small target detection method combining local gravity and horizontal line connectivity. Background Art

[0002] Infrared small target detection is a key remote sensing technology that is widely used in forest monitoring, pollution source detection, crop health assessment and other fields. In infrared images, targets usually occupy a small area with few pixels, making it challenging to capture detailed features such as shape and texture. These targets are also easily occluded by the background. In addition, low-contrast, complex and dynamic backgrounds exacerbate this challenge, resulting in high false alarm rates and low detection accuracy.

[0003] By mimicking the human visual system, contrast methods use grayscale or gradient to enhance target saliency and detect small targets even against cluttered or low-contrast backgrounds, making them a reliable choice for infrared small target detection. However, when the background contains complex textures and high-frequency details, grayscale contrast methods may have difficulty suppressing the background, leading to false detections. In contrast, gradient contrast methods can capture subtle intensity changes, which helps distinguish targets from high-frequency details, resulting in better results against complex backgrounds. Despite these advantages, current gradient contrast methods often fail to exploit the magnitude and direction information in the gradient. As a result, they still face challenges such as false detection in corner regions, target-background gradient similarity, and weak target gradient features.

[0004] In short, making full use of the gradient information in infrared images to improve the detection accuracy of small infrared targets under complex backgrounds is still an urgent problem to be solved. Summary of the invention

[0005] The technical problem to be solved by the present invention is: in view of the shortcomings of the prior art, a method for detecting small infrared targets combining local gravity and horizontal line connectivity is provided, which can still ensure accurate detection in complex scenes and suppress the background while maintaining the intensity of low-contrast targets.

[0006] The technical solution adopted by the present invention to solve the above problem is: an infrared small target detection method combining local gravity and horizontal line connectivity, comprising:

[0007] Step S1: Get the original infrared image , calculate the local gravitational force of the pixels in the neighborhood on the middle pixel and ;

[0008] Step S2: Calculate the multi-directional local gravity contrast according to the local gravity obtained in step S1;

[0009] Step S3: Calculate the horizontal line connectivity contrast according to the local gravity obtained in step S1;

[0010] Step S4: fusing the multi-directional local gravity contrast obtained in step S2 and the horizontal line connectivity contrast obtained in step S3 in different regions to obtain a fused contrast;

[0011] Step S5: According to the fusion contrast obtained in step S4, the background and the target are segmented using the adaptive threshold method to obtain the final detection result.

[0012] The beneficial effects of the present invention compared with the prior art are:

[0013] (1) This paper introduces local gravity into infrared small target detection to achieve refined gradient analysis, and develops a new sigmoid function to enhance the target and reduce noise;

[0014] (2) The present invention uses the distribution characteristics of the target to design a multi-directional local gravitational contrast method. This method enhances the target saliency while eliminating the interference of strong edges;

[0015] (3) The present invention proposes a horizontal line connectivity contrast method that utilizes the spatial consistency of the target and uses the directional information in the gravity map to increase the distinction between the target and the background.

[0016] (4) The present invention proposes a strategy of fusing two contrasts in different regions, which achieves good results in both detection accuracy and execution time. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 It is a flow chart of the infrared small target detection method combining local gravity and horizontal line connectivity of the present invention;

[0018] Figure 2 It is a schematic diagram of local gravity calculation of the present invention;

[0019] Figure 3 is the neighborhood graph that measures horizontal line connectivity;

[0020] Figure 4 The following are representative infrared image examples;

[0021] Figure 5 The following are test result diagrams of the implementation method of the present invention and other methods in three scenarios. DETAILED DESCRIPTION

[0022] In order to make the objectives, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and specific implementation cases.

[0023] The invention utilizes the distribution characteristics of local gravity amplitude to enhance the target and effectively suppress strong edges. The horizontal line connectivity contrast is designed according to the spatial consistency of the target to increase the difference between the target and the background. A method for detecting small infrared targets combining local gravity and horizontal line connectivity is specifically designed. Figure 1 Flow chart of the infrared small target detection method combining multi-directional local gravity and horizontal line connectivity of the present invention. Figure 1 As shown, the method comprises the following steps:

[0024] S1. Get the original infrared image , calculate the local gravitational force of the pixels in the neighborhood on the middle pixel and . Figure 2 An example is shown in which 3 3 The distance between the surrounding pixels and the middle pixel in the neighborhood is or ,in , , the angle between the surrounding pixels and the middle pixel Pick An integer multiple of . Based on the law of universal gravitation, the 3 3 Middle pixel in the neighborhood The local gravitational force is:

[0025] ,

[0026] ,

[0027] in, and Pixels The local gravitational force of the surrounding pixels and The directional component of force, is the gravitational constant, , l , r and Pixels and pixels The distance and angle between them.

[0028] According to the above formula for calculating local gravity, the convolution template can be determined. The convolution template and the original infrared image can be convolved to obtain the local gravity of the entire image. The convolution template for calculating local gravity is:

[0029] ,

[0030] in, It is calculated The convolution mask of the local gravity in the direction, It is calculated Convolutional mask of local gravity in direction.

[0031] The local gravity is fine-tuned using the difference and change in gravity values ​​between the target and its neighborhood. To achieve this, a new sigmoid function is introduced:

[0032] ,

[0033] in, represent The local gravity value at the location, and represent the local gravity mean and standard deviation in the neighborhood, respectively.

[0034] Then, the image The calculation method of the local gravity can be expressed as:

[0035] ,

[0036] .

[0037] S2, calculate the multi-directional local gravity contrast according to the local gravity obtained in S1. and Rotate 45 degrees to get local gravity in multiple directions. According to the distribution of small targets in different directions, divide the local gravity into target areas. and neighborhood The target area includes positive and negative force areas. The specific division method is as follows:

[0038] ,

[0039] in, is the pixel position, yes The pixel position difference in direction, yes The pixel position difference in direction, , , and is a constant.

[0040] Considering that the targets in the local gravity map are still approximately Gaussian distributed and the strong edge background is continuous, the contrast in a single direction is calculated by the following formula:

[0041] ,

[0042] ,

[0043] in, and They are Direction and Direction corresponding to Angular local gravitational contrast, and They are Direction and Direction corresponding to Angular local gravity, Take 0 degrees and 45 degrees, and for Directional contrast template, and yes Direction contrast template.

[0044] Furthermore, the multi-directional local gravity contrast can be calculated as:

[0045] ,

[0046] ,

[0047] in, Set to 0.25. and They correspond to contrasts with large target intensity and large background suppression, respectively.

[0048] S3, calculate the horizontal line connectivity contrast according to the local gravity obtained in S1. First, according to and Local gravity in the direction Calculate the gravity direction:

[0049] ,

[0050] The local gravity direction of the target area converges to the center, but this distribution difference is difficult to measure directly. To this end, the concept of the horizontal line, a unit vector perpendicular to the gravity direction, is introduced, and the horizontal line connectivity contrast is designed to enhance the difference between the target and the background. The horizontal line corresponding to the gravity direction can be expressed as:

[0051] ,

[0052] in, .from Figure 3 It can be seen that the horizontal lines in the target area have obvious connectivity. Figure 3 middle, and represents the starting point of the horizontal line vector, and represents the end point of the horizontal line vector. This connectivity can be measured as follows:

[0053] ,

[0054] in, is the horizontal line connectivity contrast, in, is the outer circle area inside the Bresenham circle with radius 2, is the inner circle of the Bresenham circle with a radius of 2. The high connectivity of the target area leads to The value is large, while the background area is disordered due to the gravitational direction. Therefore, The contrast between the target and the background is further enhanced.

[0055] The horizontal line connectivity contrast in step S3 measures the spatial consistency of the gravitational direction within the target region.

[0056] S4, the fusion contrast is obtained by regionally fusing the multi-directional local gravity contrast obtained in S2 and the horizontal line connectivity contrast obtained in S3. In the multi-directional local gravity contrast, the target usually corresponds to the area with higher value. Multiplication is used to achieve greater background suppression, resulting in higher target confidence. Therefore, we will A 7 around the maximum value The region of 7 is defined as the potential target region, while the rest of the image is designated as the potential background region. This produces a target region mask and a background region mask The fused contrast that combines the multi-directional local gravity contrast with the horizontal line connectivity contrast is expressed as:

[0057] ,

[0058] in, is the potential target region fusion contrast, is the potential background area fusion contrast, calculated as:

[0059] ,

[0060] .

[0061] The fusion contrast of step S4 is determined by determining the potential target area and the potential background area. For the potential target area, the multi-directional contrast and the horizontal line connectivity contrast are fused by addition, and for the background area, the two contrasts are fused by multiplication.

[0062] S5. Based on the fusion contrast obtained in S4, the background and target are segmented using the adaptive threshold method to obtain the final detection result. The segmentation threshold is:

[0063] ,

[0064] in, and They are The mean and standard deviation of It is an adjustable parameter.

[0065] In summary, the local gravity model fuses the information from each pixel in the local area, and introduces a new sigmoid function to reduce noise and achieve fine gradient detection. The size and direction in the gradient can be used to distinguish the target from the background. Considering that the target presents different gradient characteristics in different directions, a multi-directional local gravity contrast is further proposed. This contrast uses the distribution characteristics of the local gravity amplitude to enhance the target and effectively suppress strong edges. In addition, in order to make full use of the directional information in the local gravity map, the horizontal line connectivity contrast is designed according to the spatial consistency of the target, which increases the difference between the target and the background. Finally, a regional fusion technique is proposed to weight the two contrasts to improve background suppression while maintaining the target intensity. Experimental results show that the present invention significantly improves the detection accuracy.

[0066] Figure 4 Three representative infrared small target images are listed to compare the present invention with different models. The experimental steps are as follows:

[0067] (1) Seven infrared small target detection models are selected, namely:

[0068] Model 1: Infrared Patch Image Model (IPI)

[0069] Model 2: Infrared Patch Vector Model (NIPPS)

[0070] Model 3: Singular Value Partial Sum Minimization Model (RIPT)

[0071] Model 4: Non-convex rank approximation minimization model (NRAM)

[0072] Model 5: Tensor Nuclear Norm Partial Sum Model (PSTNN)

[0073] Model 6: Self-Regularized Weighted Sparse Model (SRWS)

[0074] Model 7: Model combining multi-directional local gravity contrast and horizontal line connectivity contrast (technical model of the present invention)

[0075] (2) Figure 5 The detection results of the present invention and other methods are shown.

[0076] (3) Quantitatively evaluate the detection results. The evaluation indicators selected are: signal-to-noise ratio gain (SCRG) and background suppression factor (BSF). The results are shown in the following table:

[0077]

[0078] It can be seen from the experimental results that the technical solution provided by the present invention shows superior performance in background suppression and target enhancement, and has higher engineering application value.

[0079] The above description is only a specific implementation of the present invention, but the protection scope of the present invention is not limited thereto. Any person familiar with the technology can understand and think of any changes or substitutions within the technical scope disclosed by the present invention, which should be included in the scope of the present invention.

Claims

1. A method for detecting small infrared targets by combining local gravity and horizontal line connectivity, characterized in that This is accomplished by following these steps: Step S1: Get the original infrared image , calculate the local gravitational force of the pixels in the neighborhood on the middle pixel and ; Step S2: Calculate the multi-directional local gravity contrast according to the local gravity obtained in step S1; Step S3: Calculate the horizontal line connectivity contrast according to the local gravity obtained in step S1; Step S4: fusing the multi-directional local gravity contrast obtained in step S2 and the horizontal line connectivity contrast obtained in step S3 in different regions to obtain a fused contrast; Step S5: According to the fusion contrast obtained in step S4, the background and the target are segmented using the adaptive threshold method to obtain the final detection result, combining the infrared small target detection method with local gravity and horizontal line connectivity.

2. The infrared small target detection method combining local gravity and horizontal line connectivity according to claim 1 is characterized by: The local gravity calculation process in step S1 is: Based on the law of universal gravitation, the 3 3 The local gravitational force exerted by the surrounding pixels on the middle pixel in the neighborhood, ， ， in, and Pixels The local gravitational force of the surrounding pixels and The directional component of force, is the gravitational constant, , , r and Pixels and pixels The distance and angle between them.

3. The infrared small target detection method combining local gravity and horizontal line connectivity according to claim 1 is characterized in that: The multi-directional local gravity contrast calculation in step S2 is based on the local gravity distribution approximating the Gaussian distribution, and the calculation process includes: The contrast ratio in a single direction is calculated by the following formula: ， ， in, and They are Direction and Direction corresponding to Angular local gravitational contrast, and They are Direction and Direction corresponding to Angular local gravity, Take 0 degrees and 45 degrees, and for Directional contrast template, and yes Directional contrast template; Furthermore, the multi-directional local gravity contrast is calculated as: ， ， in, Set to 0.25, and They correspond to contrasts with large target intensity and large background suppression, respectively.

4. The infrared small target detection method combining local gravity and horizontal line connectivity according to claim 3 is characterized in that: The step S3 comprises: First, according to and Local gravity in the direction Calculate the gravity direction: ， The horizontal line corresponding to the direction of gravity is expressed as: ， in, ; Compute horizontal line connectivity contrast: ， in, is the horizontal line connectivity contrast, .

5. The infrared small target detection method combining local gravity and horizontal line connectivity according to claim 4 is characterized in that: The step S4 comprises: Suppress the background with large contrast A 7 around the maximum value The region of 7 is defined as the potential target region, while the rest of the image is designated as the potential background region, resulting in a target region mask and a background region mask , the fusion contrast combining the multi-directional local gravity contrast and the horizontal line connectivity contrast is expressed as: ， in, is the potential target region fusion contrast, is the potential background area fusion contrast, calculated as: ， 。

Images