针对火灾火焰识别算法复杂和对环境要求高的特点,提出了基于图像特征的火灾火焰识别方法.首先采用颜色模式预判断火灾图像,排除不具有火焰颜色的物体,然后采用图像灰度化和二维熵最大阈值法进行预分割,得到火焰疑似区域,再通过提取火焰纹理特征和形状特征等对火焰图像进行分析、决策,最终判断出是否有火焰产生.实验结果表明,该方法具有较高的探测率,能降低误报漏报率且抗干扰能力强.%Due to the disadvantages of flame recognition, such as the complexity of algorithm and high requirement on the environment . A method for recognizing flame based on image futures was presented. Firstly, the image was preprocessed based on the color mode, objects without flame color futures were smoothed out, then the image was pre-segmented by graying and 2-D max entropy method, and the candidate region was obtained, thirdly the texture futures and shape futures were extracted to analyze the image, and decision about whether there is flame was made. The experiment results demonstrated that the algorithm has high detection rate, can reduce the false positive rate and has strong anti-jamming capability.
展开▼
