大空间早期火灾火焰图像分割区域的特征描述与提取

摘要

It is very important to detect the early fire in large space to protect human life and property security.In this study,feature description and analysis of the segmented flame image were carried out to extract a varity of quantitative feature descriptors of flame region based on the detection results of flame video image.The static visual features of the flame image area were investigated from the aspects of color feature,texture feature and shape feature.Meanwhile,two kinds of descriptors (i.e.,the area change percentage and the shape similarity) were also introduced to describe the dynamic visual features.The experimental results show that both of the static visual features and the dynamic visual features can improve the effectiveness and robustness of flame recognition system.The relevant results can provide some guidance for the subsequent fire pattem recognition and flame spatial orientation,which lead to reduce the false alarm rate in the recognition system and improve the fire extinguishing efficiency of the system.%基于火焰视频图像的检测结果,对分割出的火焰图像进行特征描述和分析,提取出火焰候选区域的多种定量特征描绘子.首先从颜色特征、纹理特征和形状特征3个方面描述火焰影像区域的静态视觉特性,同时引入面积变化率和形体相似度2种描绘子来描述火焰区域的动态视觉特征,作为对火焰静态视觉特征的有效补充,以提高火焰识别系统的有效性和鲁棒性.相关结果可为后续火灾模式识别以及火焰空间定位提供理论基础,减少识别误报率,提高大空间灭火系统的工作效率.