Study On Point Target Detection In Infrared Images

On military,it is a core task to find and track the targets in space operations.The infrared small target detection technology has a direct impact on the ability of detection and tracking.Therefore,the detection of infrared small targets has been one of the research emphases in military field.A new infrared point target detection algorithm based on RVR is proposed in this paper,with the objects of the infrared point targets in complex background and the basis of the characteristics of infrared point targets and RVR theory.Firstly,we introduce the development of infrared point small target detection technology.In addition,we analyze the research status at home and abroad and explore the critical technologies and difficulties of infrared point small target detection.Secondly,we give an overview of the main characteristics of infrared point small targets.Besides,we analyze several commonly used infrared small target detection algorithms of single frame image and summarize the advantages and defects of them respectively.Thirdly,we study the relevant theoretical knowledge,including RVR model training,model prediction and the advantages of RVR relative to SVR.Combining with the gray characteristics of infrared point target,a new algorithm that suitable for the detection of infrared point target in natural background is proposed.The RVR method is used to process the detected images with regression to get relatively smooth output image.The nonlinear kernel correlation coefficient method is selected as the post-processing method.Then we combine it with the characteristics of the images before and after regression to get preliminary detection result.In order to detect and locate the infrared point targets more accurately,we adopt the binary segmentation and local summation of pixels neighborhood as the final optimization methods.At last the final detection results are obtained.Finally,the proposed infrared point target detection algorithm that based on RVR is verified by MATLAB software platform.The test results prove that the proposed detection algorithm has validity.It can remove the interference,suppress the background and highlight the targets very well.In addition,we compare the proposed detection algorithm based on RVR with methods of Top-hat transform and SVR.The experimental result shows that the values of SNR of the final images obtained by the proposed algorithm are much larger than the other two methods.It means that the infrared point target detection algorithm based on RVR has better detection performance than other methods.