PROJECT TITLE :

Efficient Pedestrian Detection via Rectangular Features Based on a Statistical Shape Model

ABSTRACT:

Automatic pedestrian detection for advanced driver help systems (ADASs) remains a difficult task. Major reasons are dynamic and complicated backgrounds in street scenes and variations in clothing or postures of pedestrians. We tend to propose a simple nevertheless effective detector for strong pedestrian detection. Observing that pedestrians typically seem upright in video knowledge, we employ a statistical model of the upright human body in that the pinnacle, higher body, and lower body are treated as 3 distinct elements. Our main contribution is to systematically style a pool of rectangular features that are tailored to this form model. As we have a tendency to incorporate completely different sorts of low-level measurements, the resulting multimodal and multichannel Haar-like options represent characteristic differences between elements of the human body but are strong against variations in clothing or environmental settings. Our approach avoids exhaustive searches over all possible configurations of rectangular features nor will it depend on random sampling. It therefore marks a middle ground among recently revealed techniques and yields economical low-dimensional nevertheless highly discriminative features. Experimental results on the well-established INRIA, Caltech, and KITTI pedestrian knowledge sets show that our detector reaches state-of-the-art performance at low computational prices and that our features are sturdy against occlusions.