Smart Car Front Pedestrian Detection Based on Binocular Cameras Stereo Vision

Abstract

As a hot topic smart car front pedestrian detection is developed from the single camera to the binocular cameras, which is developed from the single camera to the establishment of a pedestrian detection system based on binocular vision goals, according to the geometric model of parallel imaging studies binocular camera binocular camera calibration and use the Software Development Kit API function call system Get the calibration parameters of the camera. On the basis of the visual system of parallel imaging geometry model of access to information on the depth of the analysis, establishing the foundation for stereo matching and three-dimensional reconstruction. As a similarity measure using the SAD function, to achieve stereo matching, and analyze the impact of matching results matches the template may cause. Through the experiment, to determine the cause of different sizes to match the template matching results have great differences, and by analyzing the results to determine the size of the template used herein match.