Research On 3D Reconstruction Technology With Marking Points And Its Application To Dimension Measurement Of Castings

With the development of the computer vision technology and digital image processing technology, photogrammetry based on 3D reconstruction technology is widely used in the fields like architecture, medical engineering, archeology, industry product detection and production process control, because of its remarkable advantages, such as high effectiveness, non-contact, excellent environmental adaptability and reliable accuracy.By traditional methods, the tasks of casting dimension measurement are usually carried out by the square rulers, shifting gauges and some other measuring equipment. However, for the measurement of big or huge castings, flexible measuring arm, Gantry three-coordinate measuring machine or some other large measuring equipment are must be used. Obviously, those equipment are not only operating-complex and inefficient but very expensive. This dissertation is just sourcing from these problems. Based on the combination of photogrammetry with 3D reconstruction technology and the practical application demand, an economical, efficient and reliable dimension measurement method is proposed in this dissertation. Also a casting dimension measurement system is developed, which is based on 3D reconstruction technology with marking points and pinhole camera imaging model, using the tool of MFC and the OpenGL technique.Firstly, in this dissertation, projective geometry and camera imaging model are theoretically analyzed and the 3D reconstruction system scheme is approximately designed by the same time. Secondly, under the conditions of the lab, the interior and external parameters of digital camera are respectively modeled and calibrated. Especially for the external parameters, a right-angle three points calibration method is utilized and, furthermore, optimized. Based on the calibration results, the camera projection matrix can be solved efficiently and accurately. Thirdly, the function of feature point extracting which can run against image noise interference is realized by using Harris corner detector with gauss-filter and non-maximum neighborhood suppression. Then, the function of feature point matching is realized by the relationship of epipolar geometry and the value of fundamental matrix. On the basis of all the work above, the task of 3D reconstruction of marking points and space lines is finally accomplished. To be a necessary part of the entire system, the post-processing module, like the function of showing results, is also developed. The detailed system operation process is introduced by applying to two specific examples at the end of this dissertation.Practical applications show that the 3D reconstruction system studied and developed by this dissertation can fulfill the casting dimension measuring tasks conveniently and efficiently. Meanwhile, it is also able to achieve the precision requirement for actual industry production. Together with the advantages of low-cost and outstanding environmental adaptability, it has great research and practical value.