作者: S. Kumar , C. Micheloni , C. Piciarelli , G.L. Foresti
DOI: 10.1109/AVSS.2009.52
关键词:
摘要: In this paper, a stereo framework for robust real time localization of objects using network’s camera pairs is presented. The system contains combination static and pan-tilt-zoom (PTZ) cameras instead traditional dual head mounted cameras. proposed novelty consists in applying vision to heterogeneous belonging video-surveillance network. First, look-up-table (LUT) built with the rectification transformations computed some predefined pan tilt values. Then, LUT used compute by means neural networks any arbitrary settings. Different zoom levels are compensated resizing images according their focal ratio zero padding. Localization object made its 3D position information obtained modified concept. Experimental results presented moving parking lot scenario.