中文摘要：

针对依靠单一色彩空间的视频检测森林防火算法中存在低准确率的问题以及视频检测森林防火算法中存在的高耗时问题,提出海计算模式下基于YCrCb色彩空间和火焰频率的视频森林火焰识别算法。在森林中部署的无线传感网的传感器端节点捕获视频图像;把图像的RGB色彩空间转换为YCrCb色彩空间,提取火焰疑似区域;通过计算视频中相邻两帧相关性系数,进行离散傅里叶变换得到火焰疑似区域的相关性系数频率,以此判断视频中是否存在火焰。该算法可以在传感器端节点实现火焰的判断。实验结果表明,该算法在准确率略微下降的情况下,大幅降低了能耗。

英文摘要：

A video forest fire recognition algorithm using color feature and frequency was proposed.Sensor nodes capture the video first.Then,the video was analyzed,the color space RGB was transformed to the YCrCb,the suspected flame area was extracted,the relative coefficient between two frames in the video was calculated,the time-domain plot of the relative coefficient was reached,and the time-domain plot was transformed to the frequency-domain plot using discrete Fourier transform（DFT）.The flame frequency was obtained through steps above.Experimental results show that the flame is judged by the frequency in this algorithm,which achieves steady accuracy and low time calculation.