A New Multi-stage Infrared Small-target Detection Method

In the modern high-tech battlefield,long-range precision strike technology has become the dominant trend of attacking methods in the war.With the development of long-range weapons,imaging system is occupying an increasingly important position.Because that infrared imaging technology has many advantages such as all-time,high imperceptibility and far effective distance,it has been widely applied in the military.In practical applications,infrared images always have very low SNR because of the far detection range.Small targets under complex background are likely to be inundated with noise which makes it difficult to detect infrared small targets.Therefore,further study and discussion on infrared small target detection is very necessary.Firstly,this paper introduces the background and significance of infrared small target detection.The research status and development trend of the problem are discussed and the background,noise and the target area of infrared small target images are emphasized.The article describes small target detection common technique in detail,such as background suppression,target enhancement technology and target segmentation technology.In addition,the common techniques of classical algorithms are analyzed.Secondly,we aim at the shortcomings of the current pretreatment methods for infrared small target background suppression in the complex background.Based on the analysis and implementation of the classical saliency detection algorithm,this paper applies the SR saliency detection algorithm to the infrared small target images.After analyzing the characteristics of the different regions of infrared small target image saliency map,a new background suppression method of weighted local entropy is proposed to perfect the saliency map.This method can make full use of the high brightness and singularity of the small target,greatly improves the SCR of the image,and effectively suppresses the complex background.Then,starting from the image scale decomposition algorithm,this paper introduces pyramid transform algorithm.The Gaussian pyramid transform is chosen in the detection because the algorithm can keep the image visual features mostly.The paper proves the rationality of pyramid transform algorithm in infrared small target detection through analysis the peak curve near the small targets.To calculate the relationship between the number of decomposition layers and the SCR of images by quantitative,we draw the conclusion that it can get best performance to keep the target and suppress the background when the number of decomposition layers is three.Finally,a new method is proposed for infrared small-target detection based on multi-stage analysis under complex backgrounds.This method consists of three steps.First,Gaussian pyramid transform is employed to achieve multi-scales.Furthermore,saliency detection is performed in each scale image using Spectral Residual approach,and process the saliency map with weighted local entropy method.Thirdly,the processed images are fused with different coefficients to get the small targets with common threshold method.This is inspired by the fact that a small target can often draw attention of human eyes and appears as a small spot in infrared images.Experimental results demonstrate that the algorithm proposed is more accurate compared with some existing methods,such as Morphological filter,Model-based method,and Max-Median filter.And the method proposed has a high detection rate and low false alarm rate.It provides a good thought for infrared small target detection algorithm.