Research On Non-uniformity Correction Technology Of Airborne Infrared Small Target Detection System

The airborne infrared search and tracking system(Infrared Search and Track,IRST)benefits from its night vision,anti-concealment and fog penetration capabilities,and has been widely used in military fields such as visual surveillance and missile guidance.Usually the detected target is very small,and it appears as a dark spot target on the focal plane.During the imaging process,it is easily affected by factors such as atmospheric radiation,complex sky background,and infrared system noise.As a result,the radiation intensity of the background noise in the infrared image is higher than that of the point target,causing the point target to be submerged in the background or clutter appears,causing false alarms.Non-uniform noise is the main source of noise in long-wave infrared imaging systems,and it is also the main reason that restricts point target detection to the background limit.Therefore,how to reduce the non-uniform noise of IRST is an urgent problem.At present,non-uniformity correction(NUC)methods are mainly divided into two categories: calibration-based and scene-based methods.Calibration-based methods can be divided into one-point,two-point and multi-point NUC.Although this type of method is simple and easy to implement,the non-uniformity caused by the nonlinearity of the detector response and drift cannot be corrected in real time.There are two types of scene-based NUC methods:(1)Statistics-based method.This type of method relies on data statistical assumptions of pixel radiation in time or space,and completes the NUC by constantly updating and correcting parameters.The disadvantage of the calibration process is that certain application scenarios are difficult to meet its premise and assumptions,and ghosting is easy to occur.(2)Registration-based method.This type of method assumes that different pixels have the same response to the same scene within a certain period of time.The inter-frame movement of the image needs to be estimated,and the algorithm calculation and storage space needed are large.Errors are easy to accumulate and transfer,and it is difficult to realize in engineering.Based on the in-depth study of the laboratory calibration method,this paper analyzes the influence of the detector response nonlinearity,random noise,optical lens and reference temperature selection on the correction effect of the two-point calibration method,and analyzes the results through experiments The verification results show that:(1)Both response nonlinearity and random noise bring non-uniformity correction errors to the infrared detection system.The response nonlinearity can be solved by the multipoint method,but it increases the complexity of engineering applications;Random noise indicates the system's detection non-uniformity background limit.Targets below this limit will be submerged in this background limit;(2)The optical lens will first reduce the signal transmission efficiency of the detection system,and secondly,the radiation and transmittance non-uniformity of the lens itself will introduce additional noise,which will seriously affect the correction effect of the two-point method;(3)When selecting two reference temperature points,you should first ensure that the two reference points have temperature points with a certain span,and then the closer the point to be corrected is to the two reference points,the better the correction effect;(4)The phenomenon of detector drifting over time has largely caused the deterioration of the two-point calibration method,and even the calibration failure for actual sky scenes.This article focuses on the influence of long-wave infrared detector drift on the effect of non-uniformity correction,an adaptive real-time detector drift compensation method based on sky background is proposed.This method can adaptively select the sky background as the reference radiation source to correct the scene.Experiments have verified that this method has a good detector drift compensation effect on sky background.Compared with the two-point calibration correction method,the method in this paper can reduce the neighborhood standard deviation from 60 to 4.9 after the same sequence image of the reference source is corrected.The reference source is different The neighborhood standard deviation of sequence images is reduced from 60 to 10,which effectively reduces the non-uniform noise caused by detector drift.Aiming at the drawbacks of the two-point calibration method and the algorithm complexity of the scene-based NUC method,a more widely applicable scene-based NUC method based on the "ratio-median method" of adjacent pixels is proposed.This method is based on the assumption of the gray values of the neighboring pixels are consistent,by calculating the gray value ratio of the neighboring pixels pixel by pixel,frame by frame,and selecting the median value of the ratio,and calculating the correction coefficient matrix recursively.Experimental results show that compared with the two-point calibration correction method,this method has a good effect in reducing the non-uniformity of target detection images,and can increase the detection distance of small infrared targets by 1.2-7.7 times in different seasons and weather conditions.The method has good real-time performance and can be carried out at any time during the flight without stopping the imaging process.In addition,the method also greatly reduces the amount of NUC algorithms for scenes in terms of the number of parameters,the algorithm process,and the amount of input data required.This makes it easier to be applied in engineering.