视频检测烟雾的研究现状

检测烟雾可以预警火灾，通过视觉监控烟雾比其它方式监控范围更广、反应更灵敏、对环境要求更低。但是目前的烟雾检测算法，无论是利用单一的色彩、纹理、形状、飘动性、闪烁以及频率等特征，或者要求满足多样特征或者采用支持向量机、神经网络、Bayesian等分类器的方法，都无法保证判断的准确性、适应性和快速性。本文综述了各种烟雾检测的方法后，认为要得到更加可靠的检测效果，一方面需要更加本质的烟雾特征，另一方面要对已有的特征进行更深入的实验验证，同时也要有更全面的样本视频数据库和算法评价标准。

State-of-art of video based smoke detection algorithms

Fire alarming, which is crucial to minimizing damage and saving lives, has no latency if it is based smoke detection. Video is a volume sensor covering an area larger than a point sensor and it is sensitive to the environment changes. Current smoke detection algorithms, whether it is based on characteristics such as chrominace, texture, shape, flutter, flicker, spatial frequency, and temporal frequency, or a composite classifier such s support vector machine, neural network, are generally difficult to make accurate, fast and robust judgment on fire. This survey reviews the state-of-art of smoke detection algorithms. We also propose three directions along the line of robust data processing. The success of the smoke detection based on visual information should be consolidated by a rigorous understanding of existing characteristics, preparing a common test database for algorithm validation and comparison, and creating new criteria for evaluating diversities of the algorithms.