Obstacle size and proximity detection using stereo images for agile aerial robots

Abstract

For autonomous flight in adverse environment, real time obstacle information is very important. There are many sensors like LiDAR, SAR which gives pretty accurate obstacle data. However, using such devices onboard is quite expensive in terms of power and operation cost. This paper introduces a technique to find the size and distance of obstacle using stereo images by correlating the physical size to in-pixel size. This technique allows agile UAVs to fly in un-cooperative environments without any heavy sensor.