Infrared Dim Small Target Detection Under Complex Background

With the development of infrared technology,its application field covers military,medical,communication and so on.Because infrared images are vulnerable to bad weather such as rain,fog and haze,infrared targets have a small area in the image and lack texture information.Therefore,infrared dim target detection under complex background has always been a research hotspot and difficulty in the image field.In order to achieve accurate target detection,this paper studies the difference between targets and backgrounds based on the mechanism of human visual attention,and extracts the saliency of targets by using information such as information entropy,gray level,gradient and frequency domain and so on,and proposes three detection algorithms for dim targets under complex background.The main work and achievements of this paper are as follows: 1.The infrared background occupies all areas of the infrared image except the target,effectively suppressing the background can improve the accuracy of the target detection.In this paper,a saliency detection algorithm based on background suppression and multi-scale local entropy is proposed.For the first time,the Guided Filter is applied to the infrared background prediction.The predicted image retains the texture features of the image well,and the image after background suppression is obtained after difference with the original image.Secondly,the information entropy characteristics of infrared targets are studied.Since the traditional local entropy method only takes advantage of the gray-scale distribution characteristics of images,and does not make use of the gray spatial distribution characteristics of targets,a new local entropy calculation method is proposed,which can better enhance the saliency of targets by combining with multi-scale theory.Then the SUSAN filter is improved according to the gray distribution characteristics of infrared targets,which further improves the robustness of the algorithm.Finally,the simulation experiments of infrared images with different backgrounds are carried out,and compared with various algorithms,the robustness of the algorithm is proved.2.The infrared image is a single channel image,and the most direct information of target is grayscale information.The target is Gaussian-like distribution in space.It has isotropic characteristics,and the background is anisotropic.Taking advantage of this difference,this paper proposes a saliency detection algorithm based on multi-channel improved DoG filter.Aiming at the problem that the traditional DoG filter lacks directional filtering capability,the filter operator is improved by adding angle information.The improved filter operator has local directivity,and then performs multi-channel and multi-directional filtering,which can well suppress various backgrounds.At the same time,aiming at the characteristics of Gaussian-like distribution of targets and strip-like clutters,a method to enhance the saliency of target is proposed by calculating the mean and variance of gray.Experimental analysis shows that the algorithm has good robustness and accuracy.3.Based on the human visual attention mechanism,the saliency of the target is the difference between the target and the background.Using single feature to detect targets is likely to result in missed detection and false alarm.Therefore,a saliency detection algorithm based on multi-feature fusion is proposed.The algorithm studies the features of the target in grayscale,gradient and frequency domain,fuses the extracted feature saliency maps,and then detects the target through PCT transformation.The experimental comparison shows that the algorithm has the advantages of high detection accuracy,simple principle and fast operation.