Image Correction And Mosaic Of Vehicle Panorama System

Traffic accident is fiercer than a tiger. Nowadays, numbers of traffic accident every year cause particular concern. Part of the causes of traffic accidents are blind spots. The spots can threaten the safe of drivers whenever they are driving or parking. So the relevant experts put forward a concept of panoramic image. This concept is based on image processing technology, which means to processing images by some certain operations in order to reach the goals.This article mainly does the research of existing image processing technology, and then chooses appropriate arithmetic to build a new generation of vehicle-mounted image mosaic system. The system includes three parts: image preprocessing, image mosaic and image fusion.The mainly works of this article include following parts:1) Do the research of correlation techniques of image mosaic, which mainly include image correction, image projection, image registration and image fusion. Compare the differences of different arithmetic of each part, and then summarize the relative merits of each arithmetic.2) Do the research of eigenvalue detection method based on angular point, and do detailed study of Harris arithmetic.3) Do further study of traditional SIFT algorithm, and consider the requirements of real-time of vehicle-mounted system. We improve the part of feature point extraction which takes the most time of the algorithm. First of all, we should simplify the scale space pyramid. Because the images are relatively stable, we can reasonably decrease the number of steps of the scale space pyramid and guarantee the matching efficiency, which can reduce the time of images matching. Then we can decrease the description gradient of the feature points, and we can find external points more quickly, and simplify the kinds of feature points direction. Experimental proofs that after two times of simplification, the improved algorithm can reduces half of operation time, and can improve the matching efficiency in the same time.4) Do research of RANSAC algorithm, and use it to refine the matching points.5) According to the study of image correlation technology and analyze then in turns, we build a new generation of vehicle-mounted image system. This system includes three parts of image pretreatment, image mosaic and image fusion. We will choose the most befitting algorithms of every parts by researching the characteristic of vehicle-mounted images. Then we will do simulations of preferential algorithms and give specific numerical value to verify the feasibility and rationality of the algorithms. At last we use GUI of MATLAB to make a human-computer interaction interface integrating choosing algorithms, and classify the interface by the functions. Users can complete the functions in order or by themselves.