Improving Robustness of Vision Based Localization Under Dynamic Illumination
MetadataShow full item record
A dynamic light source poses significant challenges to vision based localization algorithms. There are a number of real world scenarios where dynamic illumination may be a factor, yet robustness to dynamic lighting is not demonstrated for most existing algorithms. Localization in dynamically illuminated environments is complicated by static objects casting dynamic shadows. Features may be extracted on both the static objects and their shadows, exacerbating localization error. This work investigates the application of a colour model which separates brightness from chromaticity to eliminate features and matches that may be caused by dynamic illumination. The colour model is applied in two novel ways. Firstly, the chromaticity distortion of a single feature is used to determine if the feature is the result of illumination alone. These features are removed before the feature matching process. Secondly, the chromaticity distortion of features matched between images is examined to determine if the monochrome based algorithm has matched them correctly. The evaluation of the techniques in a Simultaneous Localization and Mapping (SLAM) task show substantial improvements in accuracy and robustness.
Showing items related by title, author, creator and subject.
LeCras, Jared; Paxman, Jonathan; Saracik, Brad (2011)Localization in dynamically illuminated environments is often difficult due to static objects casting dynamic shadows. Feature extraction algorithms may detect both the objects and their shadows, producing conflict in ...
Le Cras, Jared R (2012)This thesis describes the development and implementation of a multisensor large scale autonomous mapping system for surveying tasks in underground mines. The hazardous nature of the underground mining industry has resulted ...
Leoputra, Wilson Suryajaya (2009)Foreground object detection is a fundamental task in computer vision with many applications in areas such as object tracking, event identification, and behavior analysis. Most conventional foreground object detection ...