A Variable Baseline Distance Stereo Vision System for Fire Localization Based on Sub-pixel Detection

Abstract

In recent years, fire-fighting robots have been widely used to replace firefighters in dangerous environments, which has attracted the attention of more and more researchers. However, accurate fire location is an important prerequisite for quick and accurate water injection into the fire. In this paper, a variable baseline distance stereo vision system based on sub-pixel detection is proposed and its effectiveness is verified by experiments. The application of sub-pixel detection method in calibrating stereo vision system can significantly improve the accuracy and stability of fire positioning. However, with the increase of fire distance, the positioning error would increase significantly, which is unavoidable. The experimental results show that the increase of the baseline distance can effectively suppress the increase of the positioning error under the baseline distances of 186mm, 249mm and 438mm. Therefore, after an appropriate baseline distance is matched, the variable baseline distance stereo vision system based on sub-pixel detection can be used for fire positioning of wider range. The system demonstrates outstanding potential of being carried on firefighting robots to perform automated firefighting missions and providing reasonable data reference for making firefighting decisions.