基于逻辑回归模型的火焰检测

Flame detection based on logistic regression model

为了提高图像型火焰检测算法的准确率,提升算法的运行效率,通过对火焰颜色和动态特征的研究,提出了一种基于逻辑回归(LR)模型的火焰识别算法。首先利用高斯混合模型提取画面中的运动目标,在HSV色彩空间内根据火焰颜色模型分割出疑似火焰区域;然后提取疑似火焰区域的颜色及动态特征;最后将特征量输入LR模型进行参数学习,建立关于火焰特征的火焰概率模型,用于火焰识别。实验结果表明,该火焰识别算法具有识别率高、误报率低、运行速度快等优点,在图像型火焰检测上有着良好的应用前景。

In order to improve the accuracy of the image-based flame detection algorithm and improve the operating efficiency of the algorithm,a flame recognition algorithm based on the logistic regression(LR)model was proposed by studying the color and dynamic characteristics of the flame.Firstly,the moving object in the picture is extracted by using the Gaussian mixture model,and the suspected flame region is segmented according to the flame color model in the HSV color space.Then the color and dynamic features of the suspected flame region are extracted.Finally,the feature parameters are input into the LR model for parameter learning and a flame probability model about flame characteristics is established for flame identification.The experimental results show that the flame recognition algorithm has the advantages of high recognition rate,low false alarm rate,fast operation speed,etc.It has a good application prospect in image flame detection.