Study On Camera Calibration And Stereo Reconstruction In Micro-Stereo Vision System

With the development of science technology, micro-machine characterized as mini shape and dimension, mini operation measure became a new high technology for people to recognize and change object world from a micro-view. For that, the author puts forward the study on double-eye micro-stereo vision, improve the efficiency and precision of mini operation.The digitization of optics micro-image is realized by the modern photoelectricity transition, computer hardware,data and image procession software. Optics image uneasy to observe is turned into the large area, high brightness, pure corlorful image in the computer screen, and it can be observed and analyzed easily ,and do kinds of disposal (such as zoom, incision, transfer, reversal...), or realize the multimedia effect of image, character, graph.The main work and innovative ideas include:The light road of micro-stereo vision is studied, and the single CCD micro-system is changed into the double CCD micro-system. The math model is studied, and the light road model of micro-stereo is put forward. The system is calibrated and computed, and the calibration parameters such as equivalent foci f, distance d between two cameras are designated. These parameters is the internal parameters in the system, and can be calibrated through different world coordinates. In this instance, the location of camera in the world coordinates is unnecessary to calibrated, and the rotary matrix and horizontal movable matrix is needless too. If the location of calibration block is changed, and several computation is processed toward different coordinates every time, the internal parameters such as UO,VO,f,d will not change. Through lots of times calibration, the result is averaged, the precise result will be obtained.After the pretreatment, an image easy to recognize and understand for computer is produced. Digital image got from pinhole model, realize the three-dimensional reestablishing, real time need is realized by combinating VC and MATLAB. the computing task is reduced and the rapidity is reproved.Pretreatment of image is studied, such as graying, smoothing, the choice of threshold value,edge detection, erosion, inflation, figure pick-up, figure following, image segmentation. The aim is to produce a breadth of image easy to recognize and understand for computer, at the same time the base is established for stereo vision. The technology of automatic choose of threshold value is used, and it make the choose more reasonable. The graying arithmetic is improved in order to pick up interest point and target easily . In a word, camera calibration and three-dimensional reestablishing are both studied deeply in this thesis, and some fruits are achieved.