Research On Infrared Weak And Small Target Detection Under Complex Background

With the rapid development of infrared technology,infrared weak and small target detection technology in a complex background is widely used in military and civilian fields.Infrared weak and small target detection technology has the advantages of all-weather work,long working distance,strong concealment and strong anti-interference ability,at the same time,it has the disadvantages of small target,weak brightness,low contrast,blurred edges,and no texture features.Therefore,infrared dim target detection has always been a difficult and hot spot in the field of image processing.In view of the above problems,this paper proposes an infrared weak and small target detection algorithm that combines multi-features and multi-directional circular gradients through investigating the research of infrared weak small target detection algorithms at home and abroad,and analyzing the characteristics of infrared images.The main work of this paper includes the following aspects:Firstly,this paper introduces the image characteristics of infrared weak and small targets,focusing on the analysis of the three elements of infrared images-target and background and noise.The target is usually gaussian distribution,and its gray value is the maximum value of the surrounding neighborhood.It appears as an isolated bright spot or bright spots.The background is usually distributed continuously over a large area.Its radiation intensity is in a gradual state.The location of the noise in the infrared image is random,and there is no stable motion state.At the same time,common infrared image enhancement algorithms are discussed,several infrared weak target detection algorithms that have appeared in recent years are briefly introduced,and two classic infrared weak target detection algorithms are simulated and analyzed.Secondly,for the image of the infrared weak target,the gradient difference between the target and the background is analyzed.The target is the maximum value of the neighborhood,which is isotropic,and the gradients in all eight directions decrease rapidly.However,the background only has gradient descent in one or several directions,and is not isotropic.By studying the target detection algorithm based on directional gradient and multi-directional circular gradient,it is proposed to optimize the multi-directional gradient algorithm based on FAST features,and for the multi-directional circular gradient method based on the gradient difference between the target and the background,A multi-feature and multi-directional circular gradient method for infrared weak and small target detection algorithm is proposed.The algorithm combines weighted local entropy,minimum local contrast and local entropy to single frame detection based on the multi-directional circular gradient method,which effectively reduces the false detection rate and improves the performance of the algorithm.Thirdly,because the target obtained from the single frame detection algorithm processing contains false targets,we use the real target to have a continuous trajectory and stable characteristics to determine the true target by match between adjacent frames,and improve algorithm performance.First,extracting the relevant features of the target area,such as size,centroid,grayscale,roundness,circularity,compactness and ORB,and then discussing the similarity measurement algorithm,and finally,an improved neighborhood decision method is proposed to filter targets between multiple frames to find the true target.Finally,the infrared video containing infrared dim targets is selected to simulate the algorithm in this paper.The results are compared with other algorithms,concluded that our algorithm has higher detection rate and lower false detection rate.