基于数字图像处理的火警监测系统研究

Study of fire detection system based on numeric image processing

提出一种基于数字图像处理及传统的火警探测器相结合的火灾报警监测系统。计算机数字图像处理系统对图像中可能存在的火焰进行识别,给出存在火焰的可能性,并与其他传统探测器的探测结果一起输入到模糊神经网络,由模糊神经网给出最终探测结果。在火焰监测子系统中,分别根据火焰的四个特征(即颜色、扰动、火焰局部形态、颜色分布),提出了快速、实用的图像处理方法。本监测系统可以弥补传统监测系统的缺陷,扩大检测范围、缩短预报时间,大大减少误报率。实验结果表明,本系统快速有效。

Presents a visual fire detection system based on image processing combined with traditional fire detection equipment. Image processing subsystem recognizes fire that occurs in video sequences, and then gives the probability of existence of any fire. There are many previous methods to detect fire mentioned in[1](Walter Phillips Ⅲ etc.),the author presents a system to detect fire based on image processing. Nevertheless, they considered fires color and temporal variation only, for detecting fire in complex condition, it seems inadequate. Therefore, in our system the recognition algorithm takes into account four aspects of fire: color, temporal variation, shape, distribution of color. The probability will be input into fuzzy-neural network as well as the outputs of other traditional sensors, the fuzzy-neural network gives the final conclusion. This system compensates limitation of traditional detection system, widen detection area, shorten alarm time, and lessen ratio of false alarm.