融合多种特征的烟雾图像检测算法

A smoke detection algorithm combined multiple properties

提出了一种融合烟雾模糊、扩散、主方向角等多种特征的烟雾检测算法。首先对图像进行运动区域提取,在对图像进行二维离散小波变换和四元数小波变换的基础上,获得了运动区域的背景模糊模型和光流场;然后再根据光流场的相位信息,计算出光流相位分布向量和主方向角;最后依据联合判别准则对背景模糊模型、光流相位分布向量和主方向角进行判别,从而判断出运动区域是否为烟雾。实验结果表明,与利用单一特征检测烟雾的算法相比,所提出的算法有效地提高了烟雾的识别率。

A smoke detection algorithm combined of blurring property, diffuse property and the property of primary orientation angle is presented. The motion region of image is picked up, and then the background blurring model and optical flow field of the motion region are obtained by performing two dimension discrete wavelet transform and quaternary wavelet transform on images. The phase distributing vector of optical flow and the primary orientation angle are computed on the phase of optical flow field. According to an union distinguishing rule judging on the background blurring model, the phase distributing vector of optical flow and the primary orientation angle, whether the motion region is smoke or not can be decided. Experimental results show, compared to the algorithm of detecting smoke using single property, the proposed method can improve the accuracy of smoke detection.