Water body detection using Two Camera Polarized Stereo Vision

Sze Hon Yan


Autonomous vehicle navigation in rainforest terrain began to receive much attention and presents many new challenges. Navigation in rain forest terrain represents a highly complex set of problems that need to be solved before it can be successfully employed. One of the major challenges is the detection of water body within the terrain. There are two types of water; 1. still as seen in puddles and small patches and; 2. runningas the case of a stream or a river. In this paper we present a technique to detect water body for an autonomous vehicle using multiple cues of sky reflection, texture, reflection and disparity pixels with the aid of a partial polarizer. We would like to show that the fusion of all these cues could detect both still and running water. In the test runs, ourframework managed to keep the false detection error rate to below 10% for both scenarios of still water condition and running water condition at a distance of within eight meters between the water body and the camera.

Full Text:



  • There are currently no refbacks.