基于独立分量分析和颜色特征的视频火焰图像分割法

Flame image segmentation method based on independent component analysis and color characteristics

早期火灾识别需要从视频图像中分割出疑似火焰区域,有效的火焰区域分割不但可以剔除杂乱背景,减小运算量,而且有助于后期特征提取和处理。通过对独立分量分析(ICA)方法应用于图像分离的研究,将其应用于视频火焰图像分割。首先将图像序列进行白化处理,接着使用ICA方法将图像中包含着独立信息的区域分离出来,最后通过颜色模型确定火焰所在的区域。实验结果表明:该方法能有效地分割出疑似火焰区域,且不易受到噪声和光线变化影响。

Early fire detection needs to segment the suspected flame area from the video image.The effective flame region segmentation can not only eliminate the background,but also has contributed to the later feature extraction and processing.According to the research of independent component analysis(ICA)method used in image separation,the ICA method is applied to video flame image segmentation.Firstly,the image sequence is whitened,then ICA method is used to separate the different regions of the image which contain independent information.Finally,the flame region is determined by color model.The experimental results show that the proposed method can effectively segment the suspected flame area,and it has less influence by noise and variation of the light.