基于帧间高频能量和相关性的烟雾检测算法研究

Algorithm of smoke detection based on the frame of high frequency energy and correlation

针对视频烟雾检测中当干扰物和烟雾极为相似时容易出现误判的问题,对基于帧间高频能量和相关性的烟雾检测进行了研究。首先通过混合高斯背景模型与三帧差法提取运动目标,再利用烟雾的颜色、扩散性、形状提取疑似目标区,对目标区域用离散小波变换获取高频能量,利用帧间高频能量和相关性对疑似区域进行烟雾判断,从而确定目标是否是烟雾并给出报警。通过MATLAB和VS2010进行仿真,结果表明,该算法可以有效提高烟雾识别检测的可靠性,减少误报。

For the accuracy of the smoke detection, it is prone to miscarriage when the distractors and smoke are very similar.The smoke detection algorithm based on the frame of high frequency energy and correlation is studied. First, the Gaussian mixture background model and three frame difference method is used to extract the moving object, then the color of the smoke, diffusivity and shape is used to extract the suspected target area. The target area is processed with discrete wavelet transform to obtain the high frequency energy. The high frequency energy and correlation between frames are used to judge suspected area, to determine if the target is smoke and give alarm information or not. It is simulated on VS2010 and MATLAB. Simulation results show that the algorithm can improve the reliability of the smoke detection, and reduce false alarm.