Design and implementation of infrared-binocular vision system for robot navigation

An infrared-binocular vision system for robot navigation in all-environments is designed and implemented. The vision system can capture video signal of two infrared cameras from two USB-interfaces by constructing a bridge between DirectShow and OpenCV. Then the number and feature-information of obstacles can be obtained via image preprocessing and segmentation. Finally, off-line calibration is used to estimate parallax in order to extract the size, distance and position information of obstacles. Experimental results indicate the vision system can meet the demands of distance measuring for robot navigation.