A method for enhancing spatio-temporal joint local contrast infrared dim small moving target detection

Abstract

The application provides a kind of enhanced spatio-temporal joint local contrast infrared weak and small moving target detection method, comprising the following steps: S1, obtaining original infrared image sequence;S2, the local contrast of each image space is calculated, S3, the original infrared image sequence of step S1 is extracted with three layers sliding window to the time domain profile line of small target, and the time domain local contrast graph is obtained;S4, the local contrast graph of space and the time domain local contrast graph are fused to obtain target saliency map;S5, weak and small target is segmented from target saliency map by adaptive threshold segmentation, and target information is output.The method of the application utilizes two efficient methods of single frame detection and time pixel profile detection, realizes faster detection speed and stronger detection real-time, and solves the problems of high performance method, high cost and long processing time.

Description

Technical Field

[0001] This invention belongs to the field of infrared image processing technology, specifically relating to a method for detecting weak moving targets in infrared images with enhanced spatiotemporal joint local contrast. Background Technology

[0002] Infrared target detection and tracking technology has been widely used and rapidly developed due to its excellent concealment, all-weather operation, and strong anti-interference capabilities. However, infrared target detection faces the following challenges in complex backgrounds: First, targets are usually small in size or far from the detector, resulting in very few pixels occupied by the target. Second, targets have low brightness due to atmospheric scattering and absorption and other effects. Third, the background scene has a low signal-to-noise ratio and dynamic changes, which interfere with target detection.

[0003] Currently, the mainstream infrared weak moving target detection methods can be summarized into four categories: single-frame detection methods, multi-frame detection methods, spatiotemporal joint methods, and deep learning methods.

[0004] Single-frame detection methods utilize information from individual images for analysis and extraction, typically offering faster real-time processing and lower computational costs. However, relying solely on a single information source leads to poor detection performance under low signal-to-noise ratio (SNR) conditions. Multi-frame detection extracts targets by comparing pixel differences between different frames. Initially, researchers introduced temporal variance filters and temporal profile models of target pixels. Based on this, temporal profile models of target pixels have been widely applied and developed. However, these algorithms are susceptible to background scene disturbances, exhibiting a high false alarm rate under bright background interference, and their reliance on a single information source results in poor performance under low SNR conditions. Spatiotemporal joint methods combine the ideas of single-frame and multi-frame detection, simultaneously extracting spatial and temporal information, thus significantly improving algorithm performance. Du Peng et al. proposed the Spatial-Temporal Local Difference Measurement (STLDM), which directly performs target detection in three-dimensional space. However, small targets often vary in size and speed, and mainstream spatiotemporal joint algorithms only compare information between adjacent frames in the temporal domain, failing to fully extract inter-frame information, leading to low stability in moving target extraction. Furthermore, performing comparisons sequentially between frames is time-consuming, and such methods are inefficient when dealing with a large number of frames. In addition, deep learning has facilitated the development of CNN-based algorithms, which typically exhibit superior detection performance. However, due to the scarcity of available datasets, as well as their high computational cost and long detection time, these algorithms face limitations, thus reducing their practicality in engineering applications.

[0005] Therefore, how to solve the problem that existing detection algorithms have a high false alarm rate when detecting weak moving targets in low signal-to-noise ratio imaging, and cannot effectively balance detection cost and detection effect, and how to provide a fast, efficient and reliable method for detecting weak moving targets are technical problems that urgently need to be solved by those skilled in the art. Summary of the Invention

[0006] The purpose of this invention is to provide an enhanced spatiotemporal joint local contrast infrared weak moving target detection method to address the problems in the prior art.

[0007] Therefore, the above-mentioned objectives of the present invention are achieved through the following technical solutions:

[0008] A method for enhancing spatiotemporal joint local contrast infrared weak moving target detection includes the following steps:

[0009] S1, acquire the original infrared image sequence;

[0010] S2, calculate the spatial local contrast of each frame of the image, including the following steps:

[0011] S2.1, use layered convolutional gradient kernels to perform convolutional filtering on the original infrared image to suppress complex backgrounds and obtain a preprocessed image;

[0012] S2.2, The local contrast of the preprocessed image processed in step S2.1 is calculated using an enhanced three-layer sliding window to obtain a spatial local contrast map;

[0013] S3, use a three-layer sliding window to extract the temporal contour line of the small target from the original infrared image sequence of step S1 to obtain a temporal local contrast map;

[0014] S4, fuse the spatial local contrast map and the temporal local contrast map to obtain the target saliency map;

[0015] S5 uses adaptive thresholding to segment weak targets from the target saliency map and outputs target information.

[0016] While adopting the above technical solutions, the present invention may also adopt or combine the following technical solutions:

[0017] As a preferred technical solution of the present invention: In step S2.2, the enhanced three-layer sliding window calculates the local contrast using the following formula:

[0018]

[0019]

[0020]

[0021] STWLCM(i,j)=STWLCM IB (i,j)×STWLCM OB (i,j) (4)

[0022] Among them, STWLCM IB (i,j) represents the enhanced local contrast between the inner neighbor cell and the target cell, and is the minimum value of the diagonal contrast of the inner neighbor cell. It represents the local contrast between the inner neighbor cell and the target cell, where the subscript IBi indicates the contrast between the target cell and the inner neighbor cell at the i-th inner neighbor cell; m T is the mean of the target cell, with the subscript T representing the target cell region; is the mean of the inner neighboring cells of the sliding window, with the subscript IB. j This represents the j-th inner neighborhood cell, and max() represents taking the maximum value; STWLCM OB It represents the local contrast between the target cell and its surrounding neighboring cells, with the subscript indicating the surrounding neighboring cells; The value is the mean of the outer neighboring cells of the sliding window, with the index OB. i Represents the i-th outer neighborhood unit; It is a constant to prevent the denominator from being 0, and STWLCM(i,j) is a local contrast map.

[0023] As a preferred technical solution of the present invention: In step S3, the time-domain three-layer window contrast formula is:

[0024]

[0025] STTLCM(i,j,k)=max[D T (i,j,k)-M P (i,j,k),0] (6)

[0026] Among them, D T (i,j,k) is the formula for a three-layer window in the time domain; M P (i,j,k) is the average gray value of the peak region of the sliding window, M O1 (i,j,k) and M O2 (i,j,k) represent the average gray values ​​of the outer window portions on both sides; ξ is a constant to prevent the denominator from being zero; STTLCM(i,j,k) is the temporal three-layer window contrast of the k-th frame.

[0027] As a preferred technical solution of the present invention: in step S4, the fusion of the local contrast map and the complexity-weighted map is performed using the following formula:

[0028] SSTLCM(i,j,k)=STWLCM(i,j,k)×STTLCM(i,j,k) (7)

[0029] Wherein, STTLCM(i,j,k) is the target saliency map, STWLCM(i,j,k) is the spatial local contrast map calculated in step S2.2, and STTLCM(i,j,k) is the temporal local contrast map obtained in step S3.

[0030] As a preferred technical solution of the present invention: in step S5, the adaptive threshold segmentation adopts the following formula:

[0031] Th SSTLCM =μ SSTLCM +K×σ SSTLCM (9)

[0032] Among them, Th SSTLCM The image segmentation threshold is represented by the subscript SSTLCM, which indicates enhanced spatiotemporal joint three-layer local contrast; μ SSTLCM and σ SSTLCM These are the average gray value and standard deviation of the enhanced spatiotemporal joint three-layer local contrast saliency map, respectively; k is the threshold segmentation parameter; when the gray value of the target saliency map pixel is greater than Th, it is the target pixel.

[0033] Compared with the prior art, the present invention has the following beneficial effects:

[0034] This invention presents an enhanced spatiotemporal joint local contrast infrared weak moving target detection method. In the spatial domain, during single-frame detection, it utilizes a three-layer window local contrast calculation to enhance the target's spatial contrast, effectively suppressing complex and strong background clutter and solving the problem of efficiently detecting weak moving targets under low signal-to-noise ratio conditions. In the temporal domain, it analyzes the target's temporal contour using three-layer window local contrast calculation, establishing a complete inter-frame contrast detection based on all frames. This addresses the issue of single-frame algorithms based on the human visual system neglecting the extraction of temporal information of moving targets. This enhanced spatiotemporal joint local contrast infrared weak moving target detection method utilizes the fusion of single-frame detection and temporal pixel contour detection, solving the problem of high false alarm rates caused by incomplete extraction of inter-frame information in comparisons between a small number of adjacent frames in mainstream spatiotemporal joint algorithms, thus enhancing the robustness of the spatiotemporal joint algorithm. Furthermore, this invention utilizes two efficient methods—single-frame detection and temporal pixel contour detection—to achieve fast detection speed and strong real-time performance, solving the problems of high computational cost and long processing time associated with high-performance methods. It has significant application prospects in the field of infrared image processing technology. Attached Figure Description

[0035] Figure 1 This invention provides a method for enhancing spatiotemporal joint local contrast infrared weak moving target detection;

[0036] Figure 2This is a schematic diagram of the layered convolutional gradient kernel in step 2.1 of this embodiment of the invention;

[0037] Figure 3 This is a schematic diagram of the three-layer window in step 2.2 of the embodiment of the present invention. In the figure, a unit is 3×3 pixels in size, where T is the target unit, IBi is the inner neighbor unit, and OBi is the outer neighbor unit.

[0038] Figure 4 This is a schematic diagram of the three-layer sliding window in step S3, which is divided into three layers: Peak represents the peak value of the target pixel, Middle represents the middle layer, and Out represents the outermost layer.

[0039] Figure 5 This is a schematic diagram illustrating the processing results of each step in this invention. Detailed Implementation

[0040] The present invention will be described in further detail with reference to the accompanying drawings and specific embodiments.

[0041] The present invention provides a method for enhancing spatiotemporal joint local contrast infrared weak moving target detection, comprising the following steps:

[0042] 1) Acquire the raw infrared sequence images;

[0043] 2) Calculate the local spatial contrast of each frame of the image. The specific steps are as follows:

[0044] 2-1) Use layered convolutional gradient kernels to perform convolutional filtering on the original infrared image to suppress complex backgrounds;

[0045] 2-2) Calculate the local contrast of the image using an enhanced three-layer sliding window to obtain a local contrast map;

[0046] 3) In the time domain, the temporal contour lines of small targets are extracted from the original infrared image using a three-layer sliding window to obtain a local temporal comparison image;

[0047] 4) The spatial domain local contrast map and the temporal domain local contrast map are fused to obtain the target saliency map;

[0048] 5) Use adaptive thresholding to segment weak targets from the target saliency map and output target information. The enhanced three-layer sliding window method described in step (2-2) calculates local contrast using the following formula:

[0049]

[0050]

[0051]

[0052] STWLCM(i,j)=STWLCM IB (i,j)×STWLCM OB (i,j) (4)

[0053] (4) In the formula, STWLCM IB (i,j) is the enhanced local contrast between the inner neighbor unit and the target unit, which is the minimum value of the diagonal contrast of the inner neighbor unit, as shown in equation (2). It is the local contrast between the inner neighbor cell and the target cell, and the subscript IBi represents the i-th inner neighbor cell, which is the target cell and the inner neighbor cell, as shown in equation (1); m T is the mean of the target cell, with the subscript T representing the target cell region; is the mean of the inner neighboring cells of the sliding window, with the subscript IB. j This represents the j-th inner neighborhood cell, and max() represents taking the maximum value; STWLCM OB It represents the local contrast between the target cell and its surrounding neighboring cells, with the subscript indicating the surrounding neighboring cells; The value is the mean of the outer neighboring cells of the sliding window, with the index OB. i This represents the i-th outer neighbor unit, and max() represents taking the maximum value; It is a constant to prevent the denominator from being 0, and is taken as 3 in this paper; STWLCM(i,j) is a local contrast map.

[0054] The temporal three-layer window contrast formula mentioned in step (3) is:

[0055]

[0056] STTLCM(i,j,k)=max[D T (i,j,k)-M P (i,j,k),0] (6)

[0057] In the formula, D T (i,j,k) is the formula for a three-layer window in the time domain; M P (i,j,k) is the average gray value of the peak region of the sliding window, M O1 (i,j,k) and M O2 (i,j,k) represent the average gray values ​​of the outer window portions on both sides; ξ is a constant to prevent the denominator from being zero; STTLCM(i,j,k) is the temporal three-layer window contrast of the k-th frame; max() indicates taking the maximum value.

[0058] The fusion of the local contrast map and the complexity-weighted map in step 5 is performed using the following formula:

[0059] SSTLCM(i,j,k)=STWLCM(i,j,k)×STTLCM(i,j,k) (7)

[0060] In the formula, STTLCM(i,j,k) is the target saliency map, STWLCM(i,j,k) is the local contrast map calculated in step 3, and STTLCM(i,j,k) is the temporal local contrast result obtained in step 4.

[0061] The adaptive threshold segmentation in step 7 uses the following formula:

[0062] Th SSTLCM =μ SSTLCM +K×σ SSTLCM (9)

[0063] In the formula, Th SSTLCM The image segmentation threshold is represented by the subscript SSTLCM, which indicates enhanced spatiotemporal joint three-layer local contrast; μ SSTLCM and σ SSTLCM These are the average gray value and standard deviation of the enhanced spatiotemporal joint three-layer local contrast saliency map, respectively; k is the threshold segmentation parameter; when the gray value of the target saliency map pixel is greater than Th, it is the target pixel.

[0064] The above specific embodiments are used to explain and illustrate the present invention, and are only preferred embodiments of the present invention, not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made to the present invention within the spirit and scope of the claims shall fall within the protection scope of the present invention.

Claims

1. A method for enhancing spatiotemporal joint local contrast infrared weak moving target detection, comprising the following steps: S1, acquire the original infrared image sequence; S2, calculate the spatial local contrast of each frame of the image, including the following steps: S2.1, use layered convolutional gradient kernels to perform convolutional filtering on the original infrared image to suppress complex backgrounds and obtain a preprocessed image; S2.2, The local contrast of the preprocessed image processed in step S2.1 is calculated using an enhanced three-layer sliding window to obtain a spatial local contrast map; S3, use a three-layer sliding window to extract the temporal contour line of the small target from the original infrared image sequence of step S1 to obtain a temporal local contrast map; S4, fuse the spatial local contrast map and the temporal local contrast map to obtain the target saliency map; S5 uses adaptive threshold segmentation to segment weak targets from the target saliency map and outputs target information; In step S2.2, the local contrast is calculated using the enhanced three-layer sliding window method, and the calculation formula is as follows: (1) (2) (3) (4) in, It is the enhanced local contrast between the inner neighbor unit and the target unit, and is the minimum value of the diagonal contrast of the inner neighbor unit; It represents the local contrast between the inner neighbor cell and the target cell, where the subscript IBi indicates the contrast between the target cell and the inner neighbor cell at the i-th inner neighbor cell; m T is the mean of the target cell, with the subscript T representing the target cell region; is the mean of the inner neighboring cells of the sliding window, with the subscript IB. j This represents the j-th inner neighborhood unit, and max() represents taking the maximum value. It represents the local contrast between the target cell and its surrounding neighboring cells, with the subscript indicating the surrounding neighboring cells; The value is the mean of the outer neighboring cells of the sliding window, with the index OB. i Represents the i-th outer neighborhood unit; It is a constant to prevent the denominator from being zero. It is a local contrast image; In step S3, the formula for the contrast ratio of the three-layer window in the time domain is: （5） （6） in, It is the formula for a three-layer window in the time domain; It is the average gray value of the peak region of the sliding window. and These represent the average grayscale values ​​of the two outer window portions, respectively. It is a constant to prevent the denominator from being zero; is the temporal three-layer window contrast of the k-th frame.

2. The method for enhancing spatiotemporal joint local contrast infrared weak moving target detection as described in claim 1, characterized in that: In step S4, the local contrast map and the complexity-weighted map are fused using the following formula: (7) in, For the target saliency map, This is the spatial local contrast map calculated in step S2.

2. This is a time-domain local comparison image obtained in step S3.

3. The method for enhancing spatiotemporal joint local contrast infrared weak moving target detection as described in claim 1, characterized in that: In step S5, the adaptive threshold segmentation uses the following formula: (9) in, The image segmentation threshold is represented by the subscript SSTLCM, which indicates enhanced spatiotemporal joint three-layer local contrast. and These are the average gray value and standard deviation of the enhanced spatiotemporal joint three-layer local contrast saliency map, respectively; k is the threshold segmentation parameter; when the gray value of the target saliency map pixel is greater than Th, it is the target pixel.

Images