Head Detection Method Based on Binocular Stereo Vision
Qiannan Zhang · 2024
Binocular stereo vision technology is a critical area of research in computer vision, with widespread applications across various domains. Addressing the need for real-time crowd flow statistics, this study presents a head detection algorithm utilizing images captured by binocular cameras. The approach involves initial image rectification followed by analysis using binocular disparity and visual target acquisition algorithms. It further identifies head contours by examining depth profiles, depth map characteristics, and geometric features of the head's circular top area obtained from binocular stereo imaging. This method employs a depth-layered approach to extract head presence information, leveraging the circular head outline as a key geometric feature for recognition through a binocular stereo vision matching algorithm. The process efficiently detects head contours under binocular vision, resulting in high accuracy and speed in head detection within complex environments.