基于视频的全天候多特征融合火焰检测算法

针对传统可视化火灾探测技术采用单独分析可见光摄像设备或红外摄像设备采集的图像进行火灾探测, 不能全天候监控火焰的问题, 提出一种全天候多特征融合的检测算法. 首先通过视频类型判断算法确定摄像头采集的每帧图像类型, 然后利用与图像类型相对应的算法提取疑似火焰区域, 最后对疑似区域进行基于DCT的火焰频率检测和帧间相关性检测, 判断火焰是否存在. 实验表明, 该检测算法解决了传统可视化探测技术不能全天候监控火焰的问题, 且能在保持火焰高检测率的同时降低误检率.

Round-the-Clock Video Flame Detection Algorithm Based on Multi-feature Fusion

A multi-feature fusion flame detection algorithm using video-type judgment algorithm was proposed.In this algorithm,the video-type of each frame was first judged.Then candidate flame region was extracted by the algorithm that corresponds image types,at last the flame frequency feature based on DCT and image correlation feature between frames in the candidate region were detected to determine the existence of flame.The traditional visualization fire detection technology is basically based on analyzing the images from infrared cameras or visible camera to detect flame.The method can not meet the requirements of round-the-clock monitoring flame in engineering application.Experiments show that the proposed detection algorithm can resolve this problem,detect flame fast and reduce the error rate.