A New Approach of Visual Clutter Analysis for Pedestrian Detection

摘要

Detection of pedestrians during driving is a complex cognitive activity. Some of the pedestrian crashes are due to driver's late or difficult perception of pedestrian's appearance. Visual clutter analysis can be used to study the factors that affect human visual search efficiency. In this research, we propose a new approach of visual clutter analysis for pedestrian detection. We design two quantitative measures: the global environment clutter score to capture the complexity of the driving environment in terms of visual search; and the local pedestrian clutter score to evaluate the search efficiency of the pedestrian in the given driving environment. Our preliminary experimental results show that our proposed measures could serve as a reasonable indicator for measuring the human vision perception of pedestrian appearance in naturalistic driving environment.