A Multi-Sensor Approach for Rapid Digitization and Data Segmentation in Reverse Engineering

Abstract

The development of a fully automated reverse engineering system currently faces two challenges; the time consuming digitization of the object due to the multi-view requirement of current industrial sensors and the conversion of copious amounts of 3-D cloud data into a compact form, compatible with CAD/CAM packages. An ideal reverse engineering system will automatically digitize the object from multiple viewpoints, segment the cloud data into constituent surface patches and generate an accurate solid model. The utilization of both a charged coupled device (CCD) camera and a 3-D laser digitizer, in the reverse engineering process, is a major step to attaining this goal. A neural network based segmentation algorithm is applied to stereo images for the location of the target object in the laser scanner work space and to generate the laser scanner path. The process automatically generates a description of an object’s surface which can be exported to a CAD/CAM package for design or manufacturing applications. [S1087-1357(00)00503-7]