摘要
针对通用模糊聚类算法进行彩色图像分割存在对初值敏感,迭代过程耗时等问题,在HSI空间结合火焰图像分布特征,采用平均值法进行初值优选,构造抑制算子和抑制因数对火焰无关区域S和I分量进行有效抑制,采用直方图聚类后进行数据融合等方式,最终实现彩色火灾图像分割。实验表明,该算法提高了彩色火灾图像分割的准确性和收敛速度。
The general fuzzy clustering algorithm for color image segmentation is sensitive to initial value and the iterative process is time-consuming. In this paper, a fire color image segmentation algorithm base on fuzzy C-means clustering in HSI space is developed. In HSI color space, with flame image distribution, the average value method is used for optimization of initial value, thereby attenuation operator and attenuation factor are constructed to attenuate the flame unrelated region on S and I component effectively. The data are clustered and integrated base on the histogram, and the color fire image segmentation is finally realized. Experiments show that this algorithm improves the accuracy and convergence rate of color fire image segmentation.
出处
《火灾科学》
CAS
CSCD
北大核心
2017年第1期49-53,共5页
Fire Safety Science
基金
绥化学院2015年科学技术研究项目"基于嵌入式视觉的火灾检测的研究"(No:K1502003)
关键词
火灾图像分割
模糊均值聚类
无关区域抑制
直方图
Fire image segmentation
Fuzzy C-means clustering
Irrelevant area attenuation
Histogram
