摘要
PTZ摄像机是一种有效监测高速公路车辆目标通行行为的设备，可以通过调整云台角度、放缩摄像机成像范围．实现对高速公路大范围一局部场景的全面观测。本文提出一种基于多视角的高速公路大范围场景全景建模方法，获取多个观测视角融合形成的大范围场景中的车辆目标前景，从而实现PTZ摄像机多个视角下对车辆目标进行连续检测定位和跟踪，解决了当前监控摄像机监测范围小、跟踪时间短、行为识别难的问题，实验结果显示本算法对高速公路通行车辆目标定位跟踪的有效性。
Pan-tilt-zoom （PTZ） camera is an effective equipment Ibr vehicle traffic behavior monitoring on the highway. By changing the pan-tilt angle and the zoom ratio, the sight range of this kind of camera can be adjusted to realize a comprehensive monitoring of the traffic situation on the highway, which could be a wide range area or the targeted area of the highway. Based on the multi-view of PTZ camera, this paper introduces a panorama modeling method lbr the large- scale area monitoring of the highway. Foreground region of the vehicle can be extracted from tile large-scale scene image which is the fusion of the images observed from different view angles. Thus, the PTZ camera can continuously detect and track the targeted vehicle from different angles of view. This method can solve many problems ill the highway monitoring, such as tile narrow monitoring range, short tracking duration and the difficulties in behavior recognition. The experimental result shows that the proposed algorithm can effectively detect and track the vehicles on the highway.
出处
《道路交通科学技术》
2017年第5期8-13,共6页
Road Traffic Science & Technology
基金
公安部科技强警基础工作专项项目“高速公路大场景下车辆动态行为可信辨识技术研究（2016GABJC28）”、公安部技术研究计划项目“高速铁路线路重要部位物联网安防系统技术研究（2016JSYJC61）”.
关键词
视频监控
车辆检测和跟踪
图像拼接
背景建模
video monitoring： vehicle detection and tracking： image mosaicking
background modeling
