基于视觉特征对比度和二维熵的火焰检测

该文提出了在开放环境下应用视觉特征对比度和二维信息熵的火焰在线检测方法.首先利用光流法获取目标的运动矢量,然后通过运动标量生成运动对比度显著性图；进一步与颜色对比度显著图融合得到一个合成的显著图,提取出疑似火焰的感兴趣区域；接着对感兴趣区域的运动矢量计算二维熵,根据二维熵的时空一致性对是否为火焰区域进行判决.该文提出的特征对比度方法大大降低了传统基于高斯金字塔的视觉注意模型的计算复杂度.实验结果表明,该文方法对环境光照变化不敏感且能够排除其它运动物体的干扰,适用于开放环境下实时火焰检测.

A Fire Detection Algorithm Based on Visual Features Contrast and 2-D Entropy

An online fire detection algorithm based on visual features contrast and 2-D entropy under the open environment is proposed.Firstly,motion vectors are calculated by using optical flow method,then the motion magnitudes are used to generate saliency map of motion contrast,which is combined with saliency map of color contrast to obtain an integrated saliency map to extract region of interests（ROIs）.Then 2-D entropy of motion vectors of ROIs is calculated.At last the judgment is done according to spatiotemporal coherency of 2-D entropy to determine whether it is real fire.The proposed method significantly decreases the computational complexity of traditional visual attention model based gauss pyramid.The experimental results show that this method is applicable to real-time flame detection in the open environment.