开放环境下视频图像的烟雾时空特征描述

Descriptions of the Spatial-temporal Features for Smoke on Video Images in Open Environments

该文针对视频烟雾在线检测中烟雾动态特征提取问题,提出了3种开放环境下烟雾的时空特征描述方法:针对烟雾动态纹理特征的时空局部二值模型、针对烟雾区域运动速度和方向的二维互信息以及针对烟雾轮廓时空波动性的信息熵。试验结果表明,这3种烟雾时空特征描述方法均能有效地鉴别烟雾和其他非烟物体。

In this paper, three describing spatial-temporal features methods were presented aiming at smoke detection online in open environment： using the spatial-temporal local binary model（LBP） to describe the dy- namic texture features of smoke, based on the 2-D mutual information by the speed and orientations of the pixel movement to describe the movement of smoke, and describing smoke contour jitter with information entropy. The experimental results showed that these methods of smoke description based on the spatial-temporal features could effectively identify the smoke and the non smoke objects.