摘要
"烟雾根"作为林火烟雾的固有属性是区分烟雾和类烟雾的重要特征,"烟雾根"的准确检测和提取对森林火灾烟雾的识别具有重要意义。为此提出了基于高斯混合模型林火"烟雾根"识别方法,首先,采用高斯混合模型GMM对林火烟雾动态区域进行前景提取,将前景区域用形态学处理为连通的区域图像;其次,采用连通区域的骨架图像,生成烟雾根节点坐标信息;最后,从骨架端点提取与融合得到烟雾根候选点。本文提出的采用高斯混合模型进行前景提取,进行"烟雾根"识别的方法,比背景差分、帧差法和vibe等动态区域提取算法具有更好的鲁棒性。
As an inherent attribute of forest fire smoke,"smoke root"is an important feature to distinguish smoke from smoke-like smoke.The accurate detection and extraction of‘smoke root’is of great significance to the identification of forest fire smoke.The paper states the fire smoke"root"recognition method based on gaussian mixture model at the first,GMM prospect of fire smoke dynamic regions are extracted by using the gaussian mixture model,and the prospect area by using morphological image processing is connected area,secondly,by using skeleton connected regions of the image,generated smoke root node coordinate information,and finally,got the candidate smoke from the skeleton endpoints root extraction and fusion.In this paper,gaussian mixture model is used for foreground extraction and"smoke root"identification,which is more robust than that of background difference,frame difference and vibe.
作者
郑鑫
高宇
陈锋
程朋乐
ZHENG Xin;GAO Yu;CHEN Feng;CHENG Peng-le(Beijing Forestry University,Beijing 100083,China)
出处
《林业和草原机械》
2020年第1期44-46,43,共4页
Forestry and Grassland Machinery
基金
国家自然科学基金青年基金:云南松树冠火发生的关键机制研究(31800549)。
作者简介
郑鑫(1995-),男,硕士研究生;通讯作者:程朋乐(1980-),男,副教授,主要从事森林火灾监测研究。
