Building and Applying of 2.5D Range Image Based on Data Fusion

Propose an algorithm to detect road area in front of the robot based on 2.5D range image to satisfy the robot to find passable road lies 20-80m in front of it. The algorithm builds a 2.5D range image by calibrating ladar and camera. Then, it clusters points in single ladar frame using KNN. Geometrical and color information were used to segment the range image and then road area was abstracted. The experiment show the algorithm can works in semi-structural environment and has ability to deal with shadow image and ladar noise.