The Monocular Vision Target-location Methods Based On Cross-ratio Invariants

The monocular vision inspection methods based on cross-ratio invariants are mainly studied in this thesis. Compared with binocular vision and multi-camera vision, monocular vision has advantages of smaller size, lower cost, relatively simpler calibration process and briefer image processing algorithm. So, it plays an important role in the applications of plane measurement and plane detection and location.The major content of chapter one is the background of techniques of vision inspection, and the research progress in domestic and overseas. The chapter two illuminates theories of monocular vision, such as camera model and geometric transformations and the invariants between them, etc. On the basis of the theories above, it introduces two methods in vision inspection. One is the analysis of two-axis optical platform based on monocular vision and the other is monocular vision inspection based on the reference points in the same field of measurement plane.For the two-axis optical platform measurement,it illustrates several basic principles of vision inspection,measurement models and simulations on error accuracy analysis of the system with MATLAB. For the same-field measurement method, it analyzes and summarizes the previous research work, including principles of system and problems occurred in testing the existing models.Chapter three presents a modified scheme of monocular vision inspection system based on cross-ratio invariants, in which the coordinates of any ground target point relative to the target center in the visual field can be measured and a four-quadrant uniform coordinates calculation formulation is achieved. By putting landmarks along axises in the measured visual field respectively, the X-axis and Y-axis of Cartesian Coordinates is determined firstly. Then another five landmarks are placed according to some rules in this system. Their world-coordinates can be measured using total station instrument. When target points fall in this system, the geometrical relationship with the known points by line will be made. After establishing equation with invariants of cross-ratio, the iteration algorithm is proposed to complish the measurement of target points. Moreover, this paper introduced the adaptive threshold algorithm on segmentation for extracting the center of landmarks and controlled condition of nonlinear equation.Finally, it summarizes and quantifies the factors that affecting the precision of the results through the experimental results and the trend of errors, providing theoretical support for improving stability of the measurement method and increasing accuracy of indication.