Pedestrian Detection Based on HOG Features and SVM Realizes Vehicle-Human Environment Interaction

Autonomous driving has to deal with human-vehicle interaction, in which one of the key tasks is to detect pedestrians. In this paper, HOG, a classical algorithm in the pedestrian detection field is used for extracting features and SVM for pedestrian classifier training. The pedestrian feature classifier is obtained through training and testing using INRIA pedestrian dataset and data acquired by autonomous vehicles. Meanwhile, we design a pedestrian detection visualization system for better application on autonomous vehicles to detecting pedestrian. This system detects the image data input by users and calls the pedestrian feature classifier that has been trained to effectively mark pedestrians.