Data Fusion Obtained from Multiple Images Aiming the Navigation of Autonomous Intelligent Vehicles in Agricultural Environment

Abstract

Visual navigation is an important research field in robotics due to low cost of cameras and the good results that these systems usually achieve. This paper presents monocular and stereo vision-based detection methods. The obstacles are detected and fused through the Dempster-Shafer theory for generating a cloud of points that contains the probability of the existence of obstacles in the environment and its distance from the autonomous vehicle. The experiments were performed in a real rural environment to evaluate and validate the approach. The proposed system has shown to be a promising approach for obstacle detection aimed at navigating an autonomous vehicle in rural and agricultural environments.