Detecting dings and dents on specular car body surfaces based on optical flow

This paper introduces a new approach to detect defects cataloged as dings and dents on car body surfaces, which is currently one of the most important issues facing quality control in the automotive industry. Using well-known optical flow algorithms and the deflectometry principle, the method proposed in this work is able to detect all kind of anomalies on specular surfaces. Hence, our method consists of two main steps: first, in the pre-processing step, light patterns projected on the body surface sweep uniformly the area of inspection, whilst a new image fusion law, based on optical flow, is used to obtain a resulting fused image holding the information of all variations suffered by the projected patterns during the sweeping process, indicating the presence of anomalies; second, a new post-processing step is proposed that avoids the need of using pre-computed reference backgrounds in order to differentiate defects from other body features such as style-lines. To that end, the image background of the resulting fused image is estimated in the first place through a method based on blurring the image according to the direction of each pixel. Afterwards, the estimated image background is used in a new subtraction law through which defects are well differentiated from other surface deformations, allowing the detection of defects in the entire illuminated area. In addition, since our approach, together with the system used, computes defects in less than 15 s, it satisfies the assembly plants time requirements. Experimental results presented in this paper are obtained from the industrial automatic quality control system QEyeTunnel employed in the production line at the Mercedes-Benz factory in Vitoria, Spain. A complete analysis of the algorithm performance will be shown here, together with several tests proving the robustness and reliability of our proposal.