Study On Data Registration And Reduction Of Reverse Engineering

As an advanced manufacturing technique, reverse engineering that can be applied to design, develop and revise products, has become an independent subject of Computer Aided Design and Computer Aided Manufacture. Data processing, the key technique of revere engineering, mainly includes data registration, data reduction and other techniques. But data registration and data reduction are the two most important. Because the quality and the numbers of 3D data point have vital effects on the efficiency and the effectiveness of model constructing, it's very significant to study on data registration and data reduction. The concept and development of reverse engineering are represented in this thesis. Data registration and data reduction are researched in detail.In this paper, the contents of this research are as follows:1. By improving the current structured-light 3D scanning system, a four-axis vision measurement system is developed, which is composed of a three-dimension motion mechanism, a structured-light sensor and a rotation table placed on the X-axis of the three dimension motion mechanism with an inclination angle between the axis of rotation table and the Z-axis of the three dimension motion mechanism. Before scanning, the object is rotated with the table for finding an optimal measuring angle, then a piece of data is scanned using the structured-light sensor. By rotating the object a few times, the periphery and the top of the object are scanned completely. At last, the measured data patches are transformed into the same coordinate system whose origin is the centre of rotated-table and Z-axis is the axis of rotation table, therefore the data patches scanned from different views can be registrated automatically.2. A five-axis structured-light vision measurement system composed of a three-coordinate measuring machine, a PH10 rotation head and a structured-light sensor is proposed to implement the automatic registration of data patches. In this system the object is fixed and the orientation of the structured-light sensor can be changed. At every fixed orientation of the sensor a region of the object is acquired and the whole object is measured by changing the orientations of sensor several times. The extrinsic parameters of the sensor are calibrated at every orientation to ensure that these data patches have the same reference.3. To resolve the problem of determining the corresponding point in ICP (Iterative Closet Point)algorithm, a new method based on "projection" and "collinearity" is presented,i.e. firstly, the corresponding sets are found by projecting the point of the overlap regions onto the surfaces which are fitted by the points in chosen areas, secondly, the corresponding points are determined on the condition that transformed point is collinear with the optical center. Moreover, based on this method and traditional ICP, a new registration algorithm called PCICP(Projection Collinearity Iterative Closet Point) is addressed.4. A non-uniform three-dimension-grid data reduction method is presented based on the octree structure, the size of grid is determined by normal vector and principal curvature.5. Based on the PCICP method and data reduction, a integrated system is developed using Microsoft Visual C++ and OpenGL(Open Graphic Library) on the WINDOWSXP operation system.The experiment results indicate that three-dimension data registration methods possess high accuracy, the data reduction method has high reduction ratio and the reducted data can reflect the characters as detail as origin data.