中文摘要：

针对室外或大空间环境下烟雾的动态特性，实现一种实时视频处理的烟雾检测算法。首先采用高斯混合模型GMM （Gaussian mixture model），利用背景差分方法实现对运动目标的提取；其次在小波域上分辨出烟雾可能出现的图像区域；再次计算目标在空间扩散的动态特性；最后采用支持向量机区分出烟雾目标。实验结果表明，该算法适应背景复杂场景，能够有效抑制环境扰动影响，快速、准确地检测视频图像中出现的烟雾。

英文摘要：

In view of the dynamic characteristics of smoke either outdoors or in a large space environment,the article realizes a real-time video process smoke detection algorithm.Firstly it uses Gaussian Mixture Model （GMM）and takes advantage of background subtraction method to realize the extraction of the moving target;secondly it distinguishes the image regions for possible smokes from wavelet domain;thirdly it computes the dynamic features of the target that spreads throughout the space;finally it introduces SVM to identify smoke target. Experiment results show that the algorithm can adapt to scenes with complicated backgrounds, effectively suppress the influence of environmental disturbance,and quickly and accurately detect smokes from video images.