Intelligent Pedestrian Detection System in Semi-dark Environment

Computer vision techniques have been widely used in various applications. In recent years, as energy efficiency have gradually become a important issues, computer vision techniques can be integrated into a smart control system that helps increase the energy efficiency by controlling the turn on of the light based on human detection. However, implement such system that detect walking human in a semi-dark environment remains a challenge. This paper proposes a novel detection technique combining movement analysis and SVM classifier to tackle this problem. This technique consist of a few steps: a statistical background model to segment moving objects as foreground, followed by an analysis model to generate pedestrian candidates based on the movement of foreground objects and lastly a SVM classifier that verify the pedestrian candidates based on the shape features.