<title>Real-time obstacle detection via subsequent 2D images</title>

Abstract

Obstacle detection is one of the most important work during the driving of autonomous land vehicles (ALV). It is the pre-requisite to drive ALV safely and precisely. A new method for obstacle detection which using the area parameters of certain obstacle (in 2D images) is introduced here. Taking use of the explicit determination of the velocities, with which area parameters of the obstacle change in subsequent images, this approach can get the depth of the obstacle quickly. In order to make the results more accurate, the Kalman filter has been used. The advantage of our method is practical and simple (no camera calibration needed), especially when it is applied on those mobile robots without high speed parallel computer systems. Together with a very simple manner that can recognize the landmarks beside the road, our detection measure can help ALV avoid obstacles and even can drive the vehicle according to the meaning of the certain landmark. This is useful for ALV running in complex environment. The approach introduced in this paper has been applied on the Labmate robots (equipped with a single CCD camera) produced by Transition Research Cooperation, Experiments' results indicate that our Labmate vehicle performed successfully in obstacle avoidance and topological map tracking.© (1995) COPYRIGHT SPIE--The International Society for Optical Engineering. Downloading of the abstract is permitted for personal use only.