Hierarchical pedestrian detection under low resolution scenario

Abstract

The pedestrian detection is a popular research field in recent years, yet the low-resolution issue is rarely discussed for yielding reasonable response time for drivers. In this study, a hierarchical pedestrian detection system is proposed to cope with this issue. In which, two independent features, orientation and magnitude, are adopted as the descriptors to detect pedestrians. Moreover, to meet the real-time requirement, the proposed Probability-based Pedestrian Mask Pre-Filtering (PPMPF) is adopted to initially filter out lots of non-pedestrian regions while retaining as more true pedestrian as possible. In addition, the concept of integral image is also adopted to simplify the calculations of the adopted features. In experimental results, some popular features such as the Haar-like feature and the edgelet feature are adopted for comparison. The results demonstrate that the proposed system offers better performance as well as high processing efficiency, and thus it can be a very competitive candidate for intelligent surveillance applications.