CAD-guided sensor planning for dimensional inspection in automotive manufacturing

Abstract

This paper addresses the vision-sensor-planning problem in part-dimensional inspection of automotive parts. First, a CAD-guided camera-planning system is developed, which utilizes the CAD information of inspected parts and a camera model to plan camera viewpoints. A recursive algorithm, which combines two existing vision-sensor-planning approaches, is developed to find feasible viewpoints. Second, to improve the performance of the eye-in-hand robot and reduce the computational cost of the robot placement problem, a new approach is developed to integrate the kinematics constraint into vision sensor planning. Experimental and simulation results demonstrate the effectiveness of our vision-sensor-planning system.