Pedestrian Detection Algorithm Based on Multi-Feature Cascade

Pedestrian detection can be applied to social security, and can effectively increase safety in traffic safety. In real life, there is a lot of complex background in the environment, such as illumination, occlusion, dress, attitude and perspective. Therefore, the pedestrian detection has been a challenging problems and a hot research topic in computer vision. In order to improve the detection speed and guarantee the detection accuracy, this paper proposed a pedestrian detection algorithm based on multi-feature cascade. The algorithm has two important points. The one, using the improved HOG feature to quickly eliminate negative samples to improve the detection speed. The other, using the cascade structure to guarantee the detection accuracy. In this paper, the experiment proves that the algorithm indeed improves the detection speed obviously, and prove that the algorithm is indeed effective in the improvement of these two important points. The algorithm is a research result which is obtained through study some classical pedestrian detection algorithms. And the algorithm shows the vitality of the classical algorithm and provides a new thinking way for researchers.