Experimental Evaluation of Monocular Vision-Based Obstacle Detection Methods

Autonomous mobile robots require efficient and robust obstacle detection for safe and effective navigation in real-world environments. This paper explores the landscape of obstacle detection techniques, particularly monocular vision methods, which have garnered significant interest due to their affordability, lightweight nature, and rapid processing capabilities. In this work, we have investigated the performances of four practical monocular vision-based methods for obstacle detection: Canny edge detection, Shi-Tomasi corner detection, Otsu thresholding, and the You Only Look Once (YOLO) object detection model. We have evaluated these methods across diverse scenarios, considering variations in environmental conditions, camera placement, and obstacle types. The findings highlight the strengths and weaknesses of each approach, offering valuable insights for selecting appropriate techniques for robust obstacle detection.