Development of Continuous Plastering by a Robotic Arm Based on Automatic Evaluation Using Machine Learning

This study aims to develop a continuous plastering system using a robot arm that collaborates with traditional plasterers. In Japanese plastering, plasterers visually evaluate surface quality and predict trowel movements. We employ an automatic evaluation method utilizing depth images and a deep convolutional neural network for large-scale image recognition. A key feature is the use of a dataset that includes various plastering defects under different image resolutions and lighting conditions. Based on the evaluation results, the robot performs additional plastering to cover unplastered areas. This system enables continuous operation under diverse conditions and reduces the workload of human plasterers.