Abstract: In view of the high false alarm rate of image fire detection based on only Hidden Markov Model, a flame recognition algorithm based on Hidden Markov Model and Support Vector Machines is proposed. Flame candidate regions of captured image are detected with the motion detection and color analysis. Then after matching the candidate regions and flame models, HMM?is?used?to?calculate?the?similarity?values between?them. The matching values are input into SVM for second recognition. The experimental results show that compared with HMM, the HMM-SVM hybrid model can effectively reduce the rate of false positives, and improve the flame recognition accuracy.

Key words: Hidden Markov Model（HMM）, Support Vector Machines（SVM）, flame recognition, matching values

摘要： 针对单一的隐马尔科夫模型在图像型火灾探测中误报率偏高的问题，提出了隐马尔科夫模型和支持向量机相结合的图像型火焰识别算法。对捕获到的图像进行运动区域检测和颜色分析，提取疑似火焰区域，利用隐马尔科夫模型计算疑似区域与火焰模型的相似度，并输入到训练好的支持向量机进行二次识别。实验结果表明，与传统单一隐马尔科夫模型相比，该方法可以有效地降低误报率，提高火焰识别准确性。

关键词: 隐马尔科夫模型, 支持向量机, 火焰识别, 匹配值