Improving Robustness of Vision Based Localization Under Dynamic Illumination

摘要: A dynamic light source poses significant challenges to vision based localization algorithms. There are a number of real world scenarios where illumination may be factor, yet robustness lighting is not demonstrated for most existing Localization in dynamically illuminated environments complicated by static objects casting shadows. Features extracted on both the and their shadows, exacerbating error. This work investigates application colour model which separates brightness from chromaticity eliminate features matches that caused illumination. The applied two novel ways. Firstly, distortion single feature used determine if result alone. These removed before matching process. Secondly, matched between images examined monochrome algorithm has them correctly. evaluation techniques Simultaneous Mapping (SLAM) task show substantial improvements accuracy robustness.