3D Surface Modeling Of Remanufactured Parts Based On Kinect

Remanufacturing guided by the theory of Product Whole Life Period, is the general designation of repairing and reusing failure part. It is an advanced, environmental friendly manufacturing technology with the advantage of high efficiency, energy saving and material saving. The first step of remanufacturing is detecting and evaluating the damage degree of the parts, traditional measurement techniques are difficult to measure irregular shaped parts, so it is essential to get the 3D model of the parts.After analyzing various three-dimensional modeling methods, a 3D modeling system of remanufactured parts is established, which uses a new kind of sensor called Kinect as the input of the system and combines with point cloud processing algorithms. Transformation matrix from depth data of the sensor to three-dimensional coordinates is studied. The two matrices calculated by camera calibration parameters and FOV(Field Of View) are consistent. With the transformation matrix, point cloud for the surface of parts is built and absolute accuracy, relative accuracy of the sensor is obtained. At the distance of 0.8m, the relative accuracy is ± 1 mm.ICP(Iterative C losest Point) algorithm is implemented by C++ language. Using ICP algorithm, local point clouds of the parts photographed from different angles form a complete point cloud model as a whole. Failure parts model and complete parts model are united in the same coordinate system by ICP algorithm and the rate of iteration convergence of ICP is studied. MC(Marching Cube) algorithm is implemented by C++ language. With MC algorithm, Point cloud model transforms into standard formats of three-dimensional surface model, such as stl, obj, etc. 3D models of various parts are established by the system above, result indicates that the dimension error of models is ± 1mm, when compared with their real size.A Boolean subtraction algorithm of point cloud is developed for remanufactured parts. The defect model of the parts is obtained from failure parts model and the complete parts model after being processed by the Boolean subtraction algorithm. Different data structures are tested to improve the efficiency of Boolean reduction algorithm, and kdtree has the highest efficiency.An optimized remanufacturing strategy is suggested that at the beginning of scanning process, K inect sensor is used for finding target area of the parts, and laser sensor is used for further scanning. Result indicates that this strategy can ensure the accuracy that dimension error of models is ± 0.1 mm, reduce time of scanning that the scanning process does not exceed 20 seconds. Meanwhile, the process is not restricted by the size of the parts.