Infrared Target Detection Based On Morphological Component Analysis

In modern high-tech war,it is important to obtain the initiative of war,so that the early warning system and the weapon system can find and track the incoming missiles,planes and warships of enemies as well as other targets at an earlier time and in a farther distance.At present,infrared dim and small target detection technology has become the key technology in the military field,such as remote sensing system,search and track system,early warning system and so on.Infrared dim and small target has low contrast and resolution,which makes it difficult to be detected and tracked.Therefore,studying the method of infrared dim and small target detection under different background conditions has great theoretical and practical significance.Moreover,the morphological component analysis based on sparse representation and morphological difference can represent the image signal accurately and efficiently.It has been widely used in image denoising,image inpainting,image separation and other fields.In this paper,the morphological component analysis theory is introduced into the infrared dim and small target detection after studying it deeply and systematically.As a result,a high performance detection algorithm is got.The main contents of this paper include the following aspects:(1)Simulation and analysis are carried out on the linear transformation enhancement,histogram equalization enhancement,neighborhood average denoising,median filtering denoising as well as wiener filter denoising algorithms in infrared image preprocessing algorithm.(2)In order to obtain the infrared dim and small target detection method,three kinds of infrared dim and small target detection algorithms based on filtering are tested and studied.Meanwhile,an analysis is also conducted on four evaluating indicators of infrared dim and small target detection algorithms that are commonly used.(3)An improved infrared dim and small target detection algorithm based on sparse representation is proposed.Firstly,the modified gaussian intensity model is adopted to construct the overcomplete infrared dim and small target dictionary;Then,the detection images are divided into blocks and sparse decomposition is carried out in the overcomplete dictionary;Finally,the sparsity concentration index is introduced to simplify the decision procedure of target,and a threshold is also set to determine whether the block contains the target.According to the comparative analysis,the algorithm can reduce the amount of calculation and ensure the detection effect.(4)An improved adaptive dictionary construction method based on morphological component analysis is proposed to detect the infrared dim target.Firstly,the dictionary learning algorithm is used to train the overcomplete dictionary and decompose it in the modified gaussian dictionary,so that the target dictionary can be filtered;Then,the detection images are divided into blocks and sparse decomposition is made in the target dictionary;Finally,the sparse representation coefficients along with the target dictionary are utilized to restructure the blocks,a map function is introduced to restructure the residual of original image block and reconstructed image block,and a threshold for residuals is also set to determine whether the block contains the target.After the modified gaussian dictionary is improved to achieve the adaptive morphological component target dictionary,the simulation results and analysis show that the detection performance of the morphological component dictionary is better than that of the modified gaussian dictionary.