基于火焰尖角特征的火灾图像识别算法

摘要

火焰尖角的变化是火灾火焰区别于常见光源最突出的特征，也是图像型火灾探测技术必然要考虑的因素之一。而图像型火灾探测技术主要是对火灾图像数据的分析与处理，提取出火灾火焰的主要特征。由于火焰尖角变化率和圆形度是早期火灾火焰的重要边缘特征，因此该算法是在对图像进行Sobel算子边缘检测和膨胀处理的基础上，通过相邻两行的像素比值和尖角与火焰面积比值来判定尖角，再根据图像相邻帧图片火焰尖角数目的变化率来确定是否为火灾。实验结果表明，该算法可以有效地提高火焰的识别准确率，减少误判。%The variation of flame sharp angles is the most prominent feature of fire flame which differentiates it from the common light source .Furthermore , it is also one of the factors to be necessarily considered in image fire detection technology .The technology of image fire detection is principally to analyse and process the fire image data and to extract the main characteristics of the fire flame .Since the variation rate of flame sharp angle and the roundness being taken are the major edge features in early fire flame , therefore the algorithm proposed in this paper is to carry out the edge detection with Sobel operator and the expansion processing .Based on that , it discriminates the sharp angle by the pixels ratio of two adjacent lines and the ratio of sharp angles and flame area and then determines whether there has the fire or not according to the variation rate of the numbers of the flame sharp angles in adjacent frame images .It's shown in experimental result that the algorithm can efficiently improve the flame recognition accuracy and reduce the misjudgement .