Accurate Stair Measurement Method for Autonomous Robot Navigation using RGB-D Camera

This paper presents a method for accurately measuring and validating stair dimensions, which are crucial for autonomous robot navigation. The method utilizes an RGB-D camera to capture point cloud data and the Point Cloud Library (PCL) to process and validate stair dimensions within the ROS framework. The system identifies the horizontal plane (tread) and vertical plane (riser) through normal estimation and plane segmentation by the region growing technique. Dimensions such as width, height, and length are calculated from these planes, with accuracy ensured through iterative refinement. The observed dimensions are validated, ensuring that all measurements fit within the expected stair range. The approach demonstrates an average detection accuracy of 97.4% for climbing up stairs and 94.49% for climbing down stairs, making it valuable for the perception system in autonomous mobile robots ascending and descending tasks.