权益分类	功能权益	普通用户	{{item.name}}会员
{{category.name}}	{{benefitItem.name}}

Wide Area Visual Surveillance Using Fixed Viewpoint Camera

使用固定视点摄像机进行广域视觉监控

基本信息

批准号：
08680401
负责人：
WADA Toshikazu
金额：
$ 1.6万
依托单位：
KYOTO UNIVERSITY (1997)Okayama University (1996)
依托单位国家：
日本
项目类别：
Grant-in-Aid for Scientific Research (C)
财政年份：
1996
资助国家：
日本
起止时间：
1996 至 1997
项目状态：
已结题

项目摘要

In this research, a target detection and tracking system has been developed by using an active camera named Fixed Viewpoint Camera (FVC in short) whose projection center coincides with the rotational center.In 1996th research term, 1.basic algorithms, 2.physical realization of FVC,3.internal camera parameter calibration, and 4.target detection and tracking system were developped. The basic algorithms consists of (1) a wide-angle image synthesis algorithm from observed images taken by FVC,(2) and a high-speed image generation algorithm at any combinations of view direction and zoom parameters from the wide-angle image. Physical realization of FVC is the method to mount cameras on the rotational stage so that its projection center coincides with the rotational center. Internal camera parameter calibration is a method to estimate the camera parameters from observed images. Based on the research mentioned above, the background subtraction is extended so as to be applied to the images taken by FVC.By using the extended background-subtraction, we developed a target detection and tracking system which detects and tracks a target by changing view direction.In 1997th research term, 5.the target detection and tracking system was extended so that the zoom parameter as well as view direction can be controlled, 6.and a real-time system was developed which estimates the target motion and system latency. To realize the zoom control, camera calibration and image generation methods were extended to cope with the zoom control. By using the extended image generation module, the system was augmented. Also, to improve the response time of the system, target-motion and system-latency estimations were introduced into the system, and a real-time system which quickly changes the view direction and zoom has been developed. By comparative experiments, it is confirmed that the developed system has the best performance than past systems in preciseness and stability.

本研究利用投影中心与旋转中心重合的主动摄像机固定视点摄像机（Fixed Viewpoint Camera，简称FVC），研制了一个目标检测与跟踪系统，在1996年的研究期中，完成了1.基本算法、2. FVC的物理实现、3.摄像机内部参数标定和检测与跟踪系统的研制。4.target基本算法包括（1）根据FVC拍摄的观察图像的广角图像合成算法，（2）以及根据广角图像在观看方向和缩放参数的任意组合下的高速图像生成算法。FVC的物理实现是将摄像机安装在旋转台上，使其投影中心与旋转中心重合的方法。摄像机内部参数标定是一种从观测图像中估计摄像机参数的方法。在上述研究的基础上，我们将背景差分法推广应用于FVC拍摄的图像，并利用该方法研制了一个目标检测与跟踪系统，该系统通过改变视角来检测和跟踪目标，在1997年的研究期中，5.对目标检测与跟踪系统进行了扩展，使其变倍参数和视角方向都可以控制，6.设计了一个实时的目标运动估计系统，实现了目标运动估计和系统延迟估计。为了实现变焦控制，扩展了摄像机标定和图像生成方法，使其能够科普变焦控制。通过使用扩展的图像生成模块，对系统进行了扩充。此外，为了提高系统的响应时间，目标运动和系统延迟的估计被引入到系统中，并已开发出一个实时系统，该系统可以快速改变视图的方向和缩放。通过对比实验，证实了所开发的系统在精度和稳定性方面优于以往的系统。