ABSTRACT
The current mainstream research of pedestrian detection is to learn from the statistical point of view, extracting features from a large number of training samples and creating a human model, then the pedestrian detection process is considered as a pattern classification problem. The advantage is to focus on learning from samples of dierent variations of the human body, with a robust feature and a reasonable choice of training samples. Combined with the structure of reasonable classification algorithm, this method can overcome many adverse conditions, such as the pedestrians diversity, the scenes diversity, and the lighting conditions diversity.
