画像処理を用いた部品ピッキングと部品キットへの嵌め込みの自動化

抄録

At aircraft production sites, a wide variety of parts is handled. To prevent mistakes in parts, the parts to be assembled are checked by fitting into molds of parts kits. The aim of this paper is to automate this work using robots. The size between the molds and the parts is tight fit, so precision work is required. In this paper, we aim to fit model-less parts into parts kits using image processing. Template matching is used to obtain the position and angle of the parts and the corresponding molds from multiple molds in the kit. Deep learning is used to estimate the grasping position and reposition the part. To correct the misalignment of the fitting position, we use a sub-camera attached to five axes of the robot. An experimental system is constructed, and the effectiveness of the proposed method is evaluated.