A Machine Vision Framework for Autonomous Inspection of Drilled Holes in CFRP Panels

This paper presents a fully autonomous framework for the inspection of drilled holes in planar carbon fiber composite panels used in the aerospace and automotive industries. The proposed framework can automatically recognize a part and extract the geometrical information from an existing library of DXF files. It then determines the location and orientation of the part with respect to the motion platform without a need for explicit programming of the part coordinate system. Visual servoing and optimal motion planning techniques are used to autonomously move the end-effector's camera to each hole to capture high resolution images. Image processing techniques are used to determine the geometrical errors and delamination factors for each hole. All of the proposed computer vision modules have been implemented in Python and OpenCV, which are open source and thus readily available to the industry. Experimental results prove that the proposed framework can efficiently and autonomously inspect drilled holes in composite panels with minimal programming required of the end-user.