Image Preprocessing for Stereoscopic Visual Odometry at Night

摘要

Cameras are a primary sensor for many robotics perception tasks. However, cameras' performance degrades in low-light conditions. In this paper, we propose a simple preprocessing pipeline to enhance captured images in low-light conditions. This pipeline is developed with visual odometry application in outdoor environments in mind. The pipeline is evaluated on a manually collected dataset and compared to the base case with no preprocessing applied.