The Research On Intelligent Non-uniformity Correction Method Of Infrared Camera

The infrared camera has a very wide range of applications in military and civil fields,but there will exist problem of non-uniformity in imaging,which is manifested as fixed pattern noise in the image and seriously affects the imaging quality.Therefore,the nonuniformity correction(NUC)is a key issue worthy of study in infrared imaging.How to improve effect of NUC in the development stage and solve the temperature drift problem in the working stage respectively are difficulties in the study of NUC.This thesis first introduce the basic knowledge of infrared imaging,including the principle of infrared imaging technology,the composition,classification and practical application of infrared imaging system.Then,we analyze the causes of non-uniformity in detail,and introduces several linear and nonlinear response models and evaluation indexes of NUC.Next,the classification and development of NUC methods at home and abroad are introduced.Then basic principles,advantages and disadvantages of two kinds of traditional NUC methods are briefly analyzed.To reducing manual operation in the production process of infrared cameras,and further improve the correction effect by fitting nonlinear response relationship,this thesis designed a framework of an intelligent camera intelligent calibration system and a neural network fitting method based on regularization constraint is proposed in the framework,which relys on the nonlinear response relationship of detectors.Neural network is used to fit the non-linear relationship of detector nonuniformity correction,Experimental data show that the method is superior to the traditional two-point method and polynomial fitting correction method.Aiming at the change of non-uniformity caused by temperature and other environmental factors in the working stage of uncooled infrared camera,a joint NUC method of multi-frame and single frame optimization based on SIFT image registration is proposed in this thesis.With the correction compensation coefficient obtained by the change of camera poses,the proposed method can reduce the response non-uniformity of detectors and the low frequency non-uniformity caused by the optical lens and the camera shell.The simulated and actual scene images prove that the method is superior to the traditional scene correction method in terms of index and imaging effect,and realizes the intelligent correction of the camera.In this thesis,the proposed algorithm for the non-uniformity problem of infrared camera production and working process has been verified by experiments,which lays a good foundation for further intelligent non-uniformity correction in hardware.