CASIA OpenIR  > 毕业生  > 硕士学位论文
Alternative TitleResearch on Image Based Fire Detection and Alarm Technology
Thesis Advisor王欣刚
Degree Grantor中国科学院研究生院
Place of Conferral中国科学院自动化研究所
Degree Discipline计算机应用技术
Keyword火灾探测 视频图像 颜色空间 混合高斯模型 报警系统 Fire Detection Video Image Color Space Gaussian Mixture Models Alarm System
Abstract图像型火灾探测报警技术利用摄像机作为探头,将被监控现场的图像信息输入计算机系统,然后利用数字图像处理技术,从图像中识别出火焰和烟雾,进而判断火灾是否发生。由于本技术具有早期报警及可应用于室外大空间等特点,近年来受到越来越多的关注,但在实际应用中,由于火灾衍生物的多变性和场景的复杂性,漏报,误报率往往比较高,系统的鲁棒性、适应性比较差。 本文将通过对火灾发生、发展过程中火焰和烟雾图像信息的分析,提出一种鲁棒性强、适用性广的火焰与烟雾探测算法。具体地,本文完成的工作和取得的研究成果主要包括以下几个方面: (1) 对火灾发生时火焰和烟雾的图像信息进行了详细分析,对可能出现的主要干扰源进行了总结和分类。针对各类干扰源的特点并结合火灾火焰和烟雾的图像信息,设计排除干扰的方法。 (2)提出了一种新的火灾火焰探测算法,该算法首先利用HSV颜色模型进行火焰区域的分割,接下来,结合火灾火焰的尖角特征进行干扰的排除。实验表明该算法时间和空间复杂度低,能实现火灾火焰的实时探测。 (3)采用基于混合高斯模型背景建模方法提取运动前景区域,并结合火灾烟雾的亮度一致性特征、动态累积特征及扩散特征这三种可视特征,实现了一种基于运动目标跟踪的火灾烟雾探测算法。实验表明该算法探测效果好,鲁棒性强,能够排除主要的干扰源,对于帧大小为320 240的视频图像,每帧的处理时间约为40ms,在精度和速度上能够满足工程需要。 (4)结合实际火灾报警系统的性能要求,深入分析了系统设计过程中的几个关键问题,实现了一种结合时间和空间信息的设计方案。实验表明该方案在不降低探测率的前提下,能够减少系统的误报率。
Other AbstractImage based fire detection and alarm technology which makes use of cameras as a probe, the scene of the image information and digital image processing technology, identify fire flame and smoke and then determine whether the fire occurred. Since the technology above can be applied to early warning and outdoor fire detection such as large space, it gets more and more attention in the recent years and good results in particular experimental environment. However, in practical applications, due to fire derivatives variability and complexity of the scene, this technology has a relatively high false alarm rate and does not have enough robustness and adaptability. In this paper, a robust and broad applicability flame and smoke detection algorithm is designed by analyzing image information of the fire development from which broke through. Finally, an algorithm, combination of temporal and spatial information to assess the fire alarm algorithm is implemented by combining the performance requirements of the fire alarm system. Specifically, as the paper work development processed, the major research results and work have been done as follow: (1) Carry out a detailed analysis of the fire flames and smoke image information, summary most of the possible major interference sources and do a classification. According to the characteristics of various types of interference sources in combination with fire flames and smoke image information, extracts effective visual features to eliminate interference and designs the whole algorithms. (2) A fire flame detection algorithm is presented, which based on the HSV color segmentation of the flame region combined with the fire flame closed angle character. The Experiments results show that the algorithm has low time and space complexity, and can achieve real-time fire flame detection. (3) A Multi-source fire smoke detection algorithm is presented. This algorithm utilizes an improved Gaussian mixture model positioning, target following as well as consistency in the brightness, motion accumulation and spread, these three effective static and dynamic characteristics. The Experimental results show that the algorithm is able to rule out the major interference source. For 320 240 size video image frame, the processing time for each frame is about 40ms which meets the accuracy and speed needs of practical engineering requirement. (4) A few key issues about system design are proposed in light of the actual performance requirements. Thro...
Other Identifier200628014629085
Document Type学位论文
Recommended Citation
GB/T 7714
魏峥. 图像型火灾探测报警技术的研究[D]. 中国科学院自动化研究所. 中国科学院研究生院,2009.
Files in This Item:
File Name/Size DocType Version Access License
CASIA_20062801462908(1642KB) 暂不开放CC BY-NC-SAApplication Full Text
Related Services
Recommend this item
Usage statistics
Export to Endnote
Google Scholar
Similar articles in Google Scholar
[魏峥]'s Articles
Baidu academic
Similar articles in Baidu academic
[魏峥]'s Articles
Bing Scholar
Similar articles in Bing Scholar
[魏峥]'s Articles
Terms of Use
No data!
Social Bookmark/Share
All comments (0)
No comment.

Items in the repository are protected by copyright, with all rights reserved, unless otherwise indicated.