Research On Infrared Small Target Detection Method Based On Principal Component Analysis

With the development of modern technology,infrared small target detection has gradually become a key technology in military,aerospace,civil surveillance and other fields.Because the small target has a long imaging distance and is susceptible to noise interference,its intensity is low,and it does not have obvious characteristics.Infrared images have many types of backgrounds,clutter and strong edges in the background will bring challenges to the detection of small targets.At present,the existing infrared small target detection methods cannot take both background suppression and target enhancement into account.In order to solve this problem,a new infrared small target detection method is proposed in this thesis.Firstly,the thesis analyzes the characteristics of infrared image target,background and noise respectively,and summarizes the characteristics of the small infrared target.The small infrared target has the characteristics of high gray value,high information entropy,and its edge is circular.In this thesis,the background of infrared image is classified and the influence of various backgrounds on detection is analyzed.On this basis,the noise types of infrared images and the causes of noise are summarized.Secondly,the thesis studies how to suppress the strong edges of the infrared image background and enhance the target,and proposes the use of structure tensor,local information entropy and multiscale local difference contrast for data fusion to suppress the strong edges of the background and enhance the target.Then,the thesis studies how to suppress the noise in the infrared image,and proposes to add the noise-patch tensor into the infrared-patch tensor model to suppress the structural noise.Finally,based on the above work,a small infrared target detection method based on denoise reweighted infrared patch-tensor is proposed.The method can detect small targets in a single frame infrared image by using tensor robust principal component analysis,and is a lowrank sparse recovery method based on principal component analysis.The method converts infrared image into an infrared-patch tensor through sliding window.The infrared-patch tensor is composed of background-patch tensor,target-patch tensor,and noise-patch tensor.The proposed method uses background weight and target weight to suppress the background and enhance the target.Alternating direction method of multipliers is used to solve the model.When the algorithm converges,the target-patch tensor is obtained,which is reconstructed into the target image,and small targets are extracted through threshold segmentation.The thesis uses a variety of complex background infrared images for comparative experiments.Compare the performance of the method proposed in this thesis with the other 8methods.The experimental results show that the method in this thesis can accurately detect the location of small targets in the infrared image and suppress the background edges well,and has strong robustness to various noises.