基于模糊聚类的图像型火灾检测

摘要

This paper proposes an image type fire detection algorithm based on fuzzy clustering. Flame color model is established. The pulsation frequency of fire-like regions is obtained by pixel motion accumulation method. The fire-like regions which are not connected but belong to the same flame are merged with supervisory and fuzzy clustering. A judgment is made by the movement characteristics of flame. Experimental results demonstrate that the algorithm has low misstatement rate and high efficiency, it can be used in most environments of fire detection.%提出一种基于模糊聚类的图像型火灾检测算法.建立火焰颜色模型,利用像素运动累积法获取疑似目标的闪烁频率,借助有监督模糊聚类法合并同一疑似火焰区域中的不连通部分,依据火焰的运动特征进行火灾判定.实验结果表明,该算法误报率较低,运行效率较高,适用于多种场景的火灾检测.