详情

原文传递 Computer Vision Traffic Sensor for Fixed and Pan-Tilt-Zoom Cameras. Final rept.

题名：	Computer Vision Traffic Sensor for Fixed and Pan-Tilt-Zoom Cameras. Final rept.
作者：	Birchfield-S.; Sarasua-W.; Kanhere-N.
关键词：	Computer-vision-traffic-sensor; Fixed-cameras; *Pan-Tilt-Zoom-cameras.;Data-collection; Traffic-parameters; Traffic-count; Traffic-speed; Classification-data; Tracking-technology.
摘要：	This project, entitled Computer Vision Traffic Sensor for Fixed and Pan-Tilt-Zoom Cameras, involved the development, installation, and evaluation of a next generation computer vision traffic sensor capable of collecting traffic parameters such as count, speed, and classification data from a single camera sensor. In contrast to existing video-based commercial systems relying upon virtual detectors (i.e. localized presence detection), the proposed sensor relies upon recent developments in computer vision to actually track vehicles through the camera's field of view. This tracking technology enables the system to overcome many of the limitations of previous approaches, such as distractions due to reflections, shadows, perspective effects (occlusions and spillover), and congestion. Some highlights of the approach include its ability to: (1) operate in a variety of lighting and environmental conditions; (2) operate with cameras that are high above the ground (50-70 feet), as well as low to the ground (25-30 feet); (3) operate with fixed or pan-tilt-zoom cameras; (4) dynamically recalibrate itself when the camera undergoes pan and tilt movements, after an initial simple six-click calibration procedure (without the delicate heuristics required for virtual detectors); (5) work in congested scenarios; and (6) produce accurate count, speed, and classification data.
报告类型：	科技报告