基于改进FCM聚类算法的火灾图像分割

Segmentation of Fire Image Based on Improved FCM Clustering Algorithm

研究火灾识别问题,火灾图像分割是火灾特征提取和识别的前提,其分割效果直接影响火灾识别的准确率。针对现有分割方法中存在的经验阈值难以确定和因彩色信息丢失导致分割不准确等问题,为了准确识别火灾图像,提出一种改进的FCM聚类的火灾图像分割方法。方法选用符合人眼视觉特性的HSI颜色空间,根据数据分布特点确定色度分量H和亮度分量I的初始聚类中心,分别在直方图特征空间进行模糊聚类处理,并利用像素的空间信息对模糊隶属度函数做了改进,最后在由两分量的模糊隶属度组成的二维特征空间上进行火灾图像分割。实验结果表明,算法可排除高亮区域的干扰,准确分割出火焰区域,为后续的火灾识别提供重要依据。

Fire image segmentation is the premise of feature extractions and recognitions,which decides the accuracy of the fire recognition.To solve the inaccurate segmentation problem that experience threshold is difficult to determine and color information is lost.A modified FCM algorithm for fire image segmentation is proposed in this paper.The algorithm is based on HSI color space,initial clustering centers of H component and I component are selected according to the data distribution.Then,clustering the components of H and I is implemented in histogram feature space,the membership function is improved with the pixel＇s spatial feature.Finally,the algorithm is performed in the two-dimensional features space which is constructed with the image pixel membership for H and I component.The results show that the new algorithm can exclude the highlight interference region and extract flame region accurately,which is very important for the fire detection.