Application of Improved RBFNN for Surface Reconstruction in a Robot Based 3D Measurement System

A robot based 3D measurement system is proposed to overcome the disadvantage that is time-consuming when using coordinate measuring machine for surface measurement. Considering the physical meanings of the Gaussian function in RBFNN, a simple and fast methodology is developed for surface reconstruction and applied in the robot based 3D measurement system. The point cloud data obtained by 3D measurement system is projected in a 2D space, then the 2D space is divided equally and the breakpoints are selected as centers for RBFNN. Therefore, the shortcoming that needs iterative computation when using FCM method to determine the centers is avoided. Furthermore, the training and test accuracy of constructed RBFNN is better than FCM technique. Finally, point cloud data collected by measurement system from a real object is used to validate the effectiveness of the presented surface reconstruction technique.