基于小波变换的森林火灾烟雾检测算法的设计

摘要

传统的森林火灾检测技术由于效率低、价格昂贵等缺点,并不适用于森林火灾探测.文中提出并设计与实现了一种基于小波变换的森林视频火灾烟雾检测方法,使用帧间差分法和质心算法提取疑似烟雾运动目标区域,然后对提取的运动目标的前景区域和背景区域进行小波能量特征提取与分析.采用国外研究机构所提供的森林火灾烟雾视频图像,对文中算法进行验证,实验结果表明,该算法能够准确地跟踪和提取疑似烟雾运动目标区域,当森林背景区域和前景区域的高频能量比值的均值大于1.2时,在3～4秒内能够检测出森林火灾烟雾,并且该算法对常见的森林树叶运动目标具有抗干扰能力.该研究工作可实现森林火灾早期预警,对森林火灾及早准确探测具有重要意义.%The traditional forest fire detection technologies don't apply to forest fire detection due to its shortcomings such as low efficiency,expensiveness.This paper proposes,designs and implements a method for forest fire smoke detection based on discrete wavelet transform.The suspicious moving object area is extracted by using frame difference method and centroid algorithm.Then the feature of wavelet high frequency energy for the foreground and background region of moving object is extracted and analyzed.In order to validate the algorithm,the paper uses the forest fire smoke video images from video database provided by the foreign research institute.The experimental results show that the algorithm can detect the forest fire smoke in 3 ～ 4 seconds,when the mean value of high frequency energy ratio for the forest background and foreground region is greater than 1.2.In addition,the algorithm has certain anti-interference ability for the common moving object of forest leaves.The research work can help realize the early warning of the forest fire,and has great significance for the accurate detection of the early forest fire.