【作者】 冯凌云；

【导师】 杨国田；

【作者基本信息】 华北电力大学（北京） ， 模式识别与智能系统， 2010， 硕士

【摘要】 近年来,随着技术进步和嵌入式图像处理系统的发展,图像型火灾探测有望成为在大型工厂、仓库、森林等大空间和野外开放空间进行火灾探测的重要手段,研究基于机器视觉的着火点自动定位问题,可以为火灾的早期发现、自动灭火提供支持,具有重要的应用价值。论文基于机器视觉定位技术,系统地讨论了图像预处理技术、摄像机模型、摄像机标定、极线校正、立体匹配、三维重建等关键问题,并在此基础上提出了着火点的单目摄像头二维定位、双目立体视觉定位方法,并在VisualC++开发平台下,以Intel公司开发的开源计算机视觉库OpenCV作为定位系统的开发工具,构建了着火点定位系统。实验证明该系统稳定性高、适应性强、实时性好、定位准确,可应用于自动灭火系统。更多还原

【Abstract】 In recent years, as technology advances and the development of embedded image processing system, image-based fire detection is expected to be an important means of fire detection in large factories, warehouses, forests and other large open and outdoor space. The research based on burning point auto-localization of machine vision, and can give a support for the early detection of fire or automatic fire-fighting, and has an important application value.The paper is based on the locating technology of machine vision, Systematically discussed the key issues such as the image pre-processing techniques, camera model, camera calibration, epipolar calibration, stereo matching, three-dimensional reconstruction, and on this basis, proposed the method of burning point localization such as Two-dimensional localization based on Monocular Camera and Binocular stereo vision localization. And we use Intel’s open source computer vision library OpenCV as a locating system development tools to achieve burning point locating system in Visual C++development platform. Through experiments it proved that the system can be applied to fire automation system with high stability, strong adaptability and good locating accuracy.更多还原