基于稀光流的疑似烟雾提取算法

Suspected smoke extraction algorithm based on thin light flow

森林火灾时有发生,不仅会造成重大的经济损失,还会影响人们的生活和生产,甚至危害人们的生命。因此,检测火灾是否发生尤为重要。随着视频监控的普及,通过视频烟雾检测技术对森林火灾进行监控变得更有研究意义。其中,疑似烟雾区域提取算法的功能对后期的分类识别有着极为重要的影响。由于常用的几种疑似烟雾区域提取算法各有局限,文章根据森林环境针对光流法进行改进,决定结合特征点检测进行稀疏光流估计,再以运动的特征点为中心分割图像块。通过运动方向是否向上降低云雾干扰,是否符合HSV颜色特征阈值等条件筛选出待分类的疑似烟雾区域。

Forest fires occur from time to time,which will not only cause great economic losses,but also affect people’s lives and production,and even endanger people’s lives.Therefore,it is particularly important to detect whether a fire occurs.With the popularity of video surveillance,video smoke detection technology for forest fire monitoring has become more meaningful.Among them,the function of the suspected smoke area extraction algorithm has a very important impact on the later classification and recognition.Due to the limitations of several commonly used algorithms for extracting suspected smoke areas,the optical flow method is improved according to the forest environment,and it is decided to combine feature point detection to estimate sparse optical flow,and then segment image blocks with moving feature points as the center.The suspected smoke areas to be classified are screened out by whether the moving direction reduces the cloud interference upwards and whether it meets the HSV color feature threshold.