Study Of A New 3-D Shape Measurement System Based On Tri-Linear CCDs

The detection problems about 3-D morphology of moving object are extensive in tunnel construction and maintenance, safety testing of high-speed trains and so on. The paper combined 3-D topography information and analysis of two-dimensional images drawing the principle of photogrammetry based on tri-line CCDs sensor Dr. Otto Hoffman proposed. A new method of 3-D shape measurement about moving object was innovated. And the paper tried to solve the problem about 3-D measurement of the bottom of high-speed trains to meet large-scale, high dynamic, high-resolution, real-time, and other special requirements. In the paper, the works were done, such as the experiments program of new 3-D shape measurement system based on tri-line CCDs, the design of hardware system about CPU processing platform, the improvements of SIFT image matching algorithm.1. The paper discussed in detail the development and status of 3-D shape measurement technology, focused on the passive triangulation method, the active triangulation method based on structured light. And the status quo of 3-D shape measurement about moving object for this project was discussed and the new 3-D measurement based on tri-line CCDs was proposed.2. The application requirements were proposed to design the measurement system of 3-D measurement about high-speed trains including a stereo camera, CPU processing platform, communication network. And the design of experimental program of system was completed finally.3. The hardware system of CPU processing platform was designed including power, reset clock, storage, and communication network. What's more, the PCB of hardware circuit about cascade, layout, routing were designed.4. Acquisition identical points from tri-line CCDs images fast and accurate was one of key process in measurement. The paper studied the common image matching algorithms, including gray matching, matching based on image understanding, feature matching, and selected SIFT algorithm which was better in ability and anti-interference. The algorithm was improved to reduce time-consuming by optimizing scale space and simplifying feature descriptor and proposing matching search strategy based imaging characteristics. Finally, the algorithm was verified by comparing experiments, and the appropriate matching threshold was selected. At last, the paper summarized thesis work and gave the prospects for further work.