Research On Adaptive Multi-scale Infrared Small Target Detection Algorithm

Infrared small target detection has always been a hot topic in the field of image processing,and it is also the core content of infrared detection systems.With the development of science and technology,infrared weak small target detection and tracking accuracy is getting higher and higher,and it has been used in industry,agriculture,transportation,and Medicine,Aerospace and many other fields are widely used.Infrared weak target detection refers to the relatively long distance from the target in the imaging system,the imaging area is small,usually only a dozen or even a few pixels,does not have a definite shape,low contrast,and does not show its own amplitude Distribution and other features,and is susceptible to background and other interference in target detection.In recent years,researchers at home and abroad have proposed many infrared weak and small target detection algorithms,which have achieved good results.Based on them,this paper starts with the background prediction estimation algorithm and the target feature extraction algorithm,which analyzes the principle and structure of the construction algorithm,and uses the prior conditions of the infrared images,establishes a new model to effectively suppress the background region.Three adaptive multi-scale infrared weak and small target detection algorithms are proposed as follows:1.An infrared target detection algorithm based on adaptive double-layer TDLMS(Two-dimensional Least Mean Square)filtering is proposed.This method first uses the structural characteristics of the two-dimensional minimum mean square filter,and constructs a two-layer filter structure,which is a background filter layer and a target extraction layer,respectively,and correspondingly introduces a background template and a target template,which can effectively filter background clutter.Interference with noise clutter,while retaining the target information as much as possible.The method proposes an adaptive step size in the process of implementation.It uses the statistical parameters of infrared images to automatically adjust the step size,iterates the optimal weight suitable for the two-dimensional minimum mean square filter,and combines weight template.Improved the effect of infrared small target detection.The experimental results show this method can adapt to the detection of infrared targets of different background types,and has better robustness and detection performance.2.A multi-scale infrared target detection algorithm based on LCME is proposed.In this method,a multi-scale window is first set to adapt to an unknown target size.For each window,a two-dimensional minimum mean square filter with adaptive parameters is used to initially suppress background clutter.Next,a new and effective local contrast mechanism based on the iterative error adjustment strategy,LCME,is defined to obtain saliency maps at different scales.The iterative error can be automatically adjusted to distinguish the target from clutter and the target size.Interference points with similar brightness effectively improve the signal-to-noise ratio of the image;optimized related statistical parameters can not only ensure its good ability to suppress the background,enhance the contrast between the target area and the background area,but also accurately identify and highlight the target Feature information.Finally,a simple decision mechanism is used to obtain the real infrared target detection result map.The experimental results show that this method is not only better than the relevant comparison methods in terms of high detection efficiency and low false alarm rate,but also has satisfactory robustness and adaptability in complex backgrounds.3.A multi-scale infrared target detection algorithm based on the region of interest block is proposed.On the basis of ensuring the target detection performance,the running speed of the target feature extraction algorithm is effectively optimized.This method first uses non-subsampled pyramid NSP to perform decomposition and transformation,extracts high-frequency information in the infrared image in layers,and then fuses,and then draws a region of interest in the fused image to reduce the target.Unnecessary calculations in detection,reducing complexity.Combined with the improved local contrast mechanism,the background is effectively suppressed,the contrast between the target area and the surrounding background area is enhanced,and the target is more prominent.Multi-scales appear in sliding windows of different scales.Finally,a simple decision mechanism is used to select the optimal scale saliency map and extract salient target feature information.Experimental results show that compared with other similar algorithms,the algorithm proposed in this chapter has better target detection performance and higher real-time performance,which is a superior method.