Trowel Marks Detection and 3D Unprojection for Interior Floor Finishing Robots

초록

Interior floor finishing is a labor-intensive work that often leads to physical strain on workers. Robotic systems are required to alleviate these challenges and enhance productivity. Manipulator-based floor finishing robot has the potential to perform highly precise tasks and improve productivity. However, trowel marks that appear on the mortar surface after finishing work caused by the manipulator’s applied force should be addressed to ensure consistency and high quality of the outcomes. Therefore, it is essential to detect irregular trowel marks on the mortar surfaces and efficiently remove the marks. This study proposes a method that detects and removes irregular trowel marks using an RGB-D sensor and a robotic arm. The scope of this study includes the detection of trowel marks and the transformation of their coordinates to provide their locations in the robot coordinate system. The proposed method uses YOLO-v8 image segmentation model to identify the trowel marks and extracted pixel-level coordinates for the corresponding area. Furthermore, a 2D-to-3D coordinate transformation based on Unprojection is applied to convert locations of the trowel marks into the robot coordinate system. This study verified the performance and effectiveness of the proposed process in the Gazebo simulation environment. The evaluation showed that the transformed coordinates had an average positional deviation of 4.3 mm along the x-axis and 2.1 mm along the y-axis. In addition, the trowel marks were effectively segmented from the mortar surface images, and corresponding trajectories were successfully generated. © 2025, Korea Institute of Construction Engineering and Management. All rights reserved.

키워드