Detecting pedestrians using patterns of motion and appearance

TLDR

Past approaches have built detectors based on appearance information, but ours is the first to combine both sources of information in a single detector. The study presents a pedestrian detection system that fuses image intensity and motion cues, offering an efficient motion representation and a state‑of‑the‑art detector for low‑resolution images in adverse weather. The detector scans two consecutive video frames, is trained with AdaBoost to combine motion and appearance cues, and extends the Viola‑Jones framework. The system achieves about 4 fps, detects pedestrians as small as 20 × 15 pixels, maintains a very low false‑positive rate, and operates effectively on low‑resolution images even in rain and snow.

Abstract

This paper describes a pedestrian detection system that integrates image intensity information with motion information. We use a detection style algorithm that scans a detector over two consecutive frames of a video sequence. The detector is trained (using AdaBoost) to take advantage of both motion and appearance information to detect a walking person. Past approaches have built detectors based on appearance information, but ours is the first to combine both sources of information in a single detector. The implementation described runs at about 4 frames/second, detects pedestrians at very small scales (as small as 20/spl times/15 pixels), and has a very low false positive rate. Our approach builds on the detection work of Viola and Jones. Novel contributions of this paper include: i) development of a representation of image motion which is extremely efficient, and ii) implementation of a state of the art pedestrian detection system which operates on low resolution images under difficult conditions (such as rain and snow).