基于视频的烟雾检测系统——运用烟雾流动模型和时空能量分析的方法

Smoke Detection System Based on Video——Through the Analysis of Smoke flow Pattern and Spatial-temporal Energy

在火灾发生的初始阶段,监测烟雾对于防止火灾至关重要。文章提出了一种基于时间特征的检测方法,即由光烟流动模式分析和时空能量分析提取的时间特征融合得到。一个特征向量是通过使用具有优选方向的Gabor滤波器组,利用纹理信息来确定烟雾的流动特征。此外,在具有时间差的图像中,应用空间频率的能量分析得到另一特征向量。最后,这些特征向量输入训练好的支持向量机(SVM)进行烟雾判别,提供准确的烟雾检测。根据实验数据可得,该算法特征提取简便,可以更快检测烟雾的发生,提高了烟雾检测效率。

In the initial stage of fire,monitoring smoke is very important for preventing fire.This paper proposes,a temporal feature extraction method,which is fused by light smoke flow pattern analysis and spatiotemporal energy analysis.A feature vector is obtained by using a Gabor filter bank with a preferred direction and using texture information.Besides,in the image with time difference,another feature vector is obtained by energy analysis of spatial frequency.Finally,these feature vectors are fed into support vector machine(SVM)to discriminate smoke and provide accurate smoke detection.According to the experimental videos,the features of algorithm are simple and easy to extract.It can detect smog quickly and improve the efficiency of smoke detection.