Sparse Representation Based Stereo Matching Algorithm And Infrared Targets Detection And Tracking

With the continuous development of the technology in digital image processing, the question of how to efficiently decompose, represent, extract the image information has become a research focus in computer vision. Computer vision learn 3-D environment from 2-D images with multi-disciplinary knowledge like image processing, psychophysics, cognitive science, neurophysiology and so on.Focus on the research content in binocular infrared radar system, this paper could be divided into two parts:first, an infrared multi-object detection and tracking algorithm in outdoor complex background; second, a sparse representation based stereo matching algorithm.Chapter 1:The background and significance of this subject, the research situation of sparse representation algorithm, and analyze some of the problems of existing algorithms.Chapter 2:Introduction of infrared targets detection and tracking methods, including preprocessing of infrared images, infrared targets detection and infrared targets tracking.Chapter 3:A multi-object detection and tracking algorithm in outdoor complex background, including three parts:a fast method for background extraction and update, a seed selection method in region growing and an optimization method for region growing.Chapter 4:Some of the basic theory of stereo vision and principles of stereo matching algorithms, including the basic principle of binocular stereo vision, camera calibration and so on.Chapter 5:A sparse representation based stereo matching algorithm. In this chapter, we shall complete the steps below:firstly extract a one-dimensional signal from a two-dimensional image, secondly establish an over-complete dictionary of atoms, thirdly, make the sparse decomposition for 1-D signal, fourthly and lastly calculate the sub-pixel disparity and the depth of the target. Chapter 6:Summary of the research work of this paper and the future prospect of the follow-up study will be carried out.