Study On Vision Mechanism And It's Application In IR Imaging Guidance System

Today, infrared imaging guidance has been one of the most precise guidance techniques. The FPA(Focus plane array) is equivalent to the retina of human, based on it, we introduce the mechanism of HVS (human vision system) to improve the imaging guidance system. This dissertation integrates the study of human vision system with the study of imaging guidance. The visual information processing mechanism of HVS is explored, Furthermore many inspired ideas and algorithms are put forward and applied to the infrared imaging guidance system which is the important module of precise guidance techniques. The main achievements of the dissertation include:1. The study of retinal image processing mechanism1) Aiming at the retinal information-processing model, the difference algorithm of recurrent lateral inhibition is presented. Based on it, the simulation of retinal information process is made. The consistency of LI and RF in detecting the edge information of scene image is pointed out. In addition, the model-constructed rules of Off-center receptive field are explored. Uniting the lateral inhibition mechanism, a novel spatial model of retina mechanism is proposed.2) Mechanism of fixation eye movement is developed; the human eyeball has continuous tiny movements so that an eye cannot be analyzed as a still camera. This dissertation focuses on the retinal image and simulates the visual information process on the viewpoint that fixation eye movement modulates the retinal image. Based on it the principles of visual hyperacuity are analyzed.3) Two kinds of mechanism to ensure eye's multi-scale ability are presented and a novel algorithm of convolution mask construction is proposed against the realization of Gauss scale space.4) For the image preprocessing work, a new image-enhanced method inspired by Fechner's law is proposed Aiming at the air target at low attitude, an efficient image segment method "inhibition of on-line background simulation" is put forward, Moreover for the object of low SNR airdrome's runway image, improved Radon Transform method is developed.2. The study of space-variant vision system based on the retino-cortical mapping1) Retino-cortical mapping algorithm is studied in detail. We propose the concept "unified one-one mapping ring". According to this ring and mapping direction we classify the mapping transformation into two sorts: forward algorithm and reverse algorithm. Moreover, enhanced by the sub-pixel interpolation, an optimized reverse algorithm to realize the non-uniformed mapping is proposed. By the non-uniformed mapping the visual information stream is modeled.2) The mathematic model's range of retino-cortical mapping is broadened. We put forward the general rules of constructing this mapping model, based on it linear model and arctangent model are inferred and verified. The reasearch of retina-like sensor is explained and the strategies of realizing rapid transformation are studied.3) Based on Log-Polar image, a novel target recognition method is developed. We explore three algorithms: one-dimension projection statistic match, double edge information match and two-dimension Log-Polar image match. Acoording to the invariant property of LPT, which the rotation and scaling of the input signal are transformed into translations that can be easily detected and compensated.4) Retino-cortical mapping is an important attribute of human visual system, which enables human visual system to have the large visual field and high acuity on gaze point simultaneously. Aiming at the infrared imaging guidance system, we deduce the applying way where the transformation parameters are determined by precise visual-field angle. In addition we analyze this application in scanning of large visual field and target tracking. Furthermore this dissertation put forward image two-module representation inspired by the large-field system and small-field system of the fly. and then a parallel model of outer-filed system and inner-field system comes into being.