WITHDRAWN: New obstacle detection method for a mobile robot based on improved disparity map processing

Abstract

• ODMC significantly reduces execution time, making it one of the fastest methods for generating disparity maps in the literature. • D-SRG effectively detects obstacles with minimal error, producing an accurate list of obstacles along with their distances from the stereovision system. The combined system achieves a processing rate of up to two frames per second (2 fps), meeting the stringent real-time navigation requirements for autonomous systems. These results demonstrate the potential of our approach for improving the efficiency and accuracy of stereovision-based obstacle detection. Stereovision is a crucial technique for vision-based obstacle detection as it provides accurate 3D information for analyzing scenes and detecting obstacles. Despite its effectiveness, existing methods often struggle with real-time performance and computational efficiency. This paper introduces a new stereovision-based obstacle detection system that integrates two novel approaches: Optimized Disparity Map Computation (ODMC), which enhances Hirschmüller’s method for faster disparity map generation, and Disparity Seeded Region Growing (D-SRG), an improved version of the SRG algorithm optimized for identifying obstacles in real-time. Our system is designed to address key challenges such as high memory consumption and slow processing speed, enabling efficient real-time processing from image acquisition to obstacle detection using a 3D camera. Experimental results show that: