The Study On High-accuracy Mosaic For Large 3D Free-form Measurement

Multi-view data mosaic is a necessary link in large 3D free-form measurement. The mosaic quality directly influences the feasibility and precision of 3D reconstruction. In this thesis, the merits and drawbacks of universal mosaic method in the world were analyzed. Both theoretic and practical researches were done based on the constrained and integral mosaic idea using unorganized cloud points as mosaic elements. Two mosaic methods were proposed respectively based on single image of auxiliary target and global control network. Some corresponding experiments were taken. It was proved that these two methods had good feasibility. They were respectively applicable for non-destructive measurement with moderate precision and precision measurement.The main work of this thesis is as follow:1. Summing-up the development of the technology of multi-view data mosaic in the world. Analysis of merits and draw backs of various mosaic methods combining application.2. Study of binocular stereo visual sensor based on laser scan. Established the scan mechanism composed of laser cluster with precise step motor and redactor. Optimized sensor structure based on analysis of measurement precision and sensor's field of view. Calibrated the camera and sensor with one-dimensional target simulating three-dimensional target. Realized rapid and accurate match of feature beam based on epipolar constraint.3. Creation a new mosaic method based on single image of auxiliary target. Completed the design, recognition and sub-pixel locating of global control points and target control points. Accomplished accurate orientation of global control points by mosaic camera simulating intersectant measurement system composed by two cameras. Realized the integral mosaic by establishing and resolving the coordinate system transfer chain from unit measurement coordinate system to auxiliary target coordinate system and to mosaic camera coordinate system and to global coordinate system based on the theory of projective geometry and perspective geometry using auxiliary target and mosaic camera as intermediary.4. Study of the mosaic method based on global control network. Completed the design, recognition and sub-pixel locating of coded and non-coded control points. Estimated and optimized the fundamental matrix, realized the rough match of non-coded control points between two images by epipolar constraint. Accurately matched all the non-coded control points among images taken by all measurement stations by three images method with relaxation match idea and numbered all the non-coded control points based on the match results. Established the ray bundle adjustment equations and resolved them to get the global control network. Mapped part of non-coded control points from unit measurement coordinate system to global control network by distance constraint.5. Experiments of measurement and mosaic for engine deck of car and rotary type object respectively based on the two mentioned above. And error analysis was done synchronously.