基于机器学习的烟雾检测算法去除固定干扰  

Remove of Fix interference Based on Machine Learning in Smoke Detection

作　　者：刘青松[1] 崔更申[1] 吴鹏[1] 李传明[1] 

机构地区：[1]桂林电子科技大学计算机科学与控制学院,广西桂林541004

出　　处：《计算机测量与控制》2015年第3期880-881,885,共3页Computer Measurement &Control

基　　金：国家自然科学基金资助项目(61063040);广西研究生教育创新计划资助项目(YCSZ2013068)

摘　　要：烟雾是早期火灾最为突出的视觉表现,因此检测火灾烟雾在日常防火应用中具有重要的意义;针对目前视频烟雾检测算法中误报率高,适应性差等特点,在实施基本烟雾检测算法并在其基础上提出基于机器学习的视频烟雾去干扰方法,能很大程度上去除固定干扰物的干扰,提高了烟雾识别的正确率;实验证明,该算法可以较好、较快地检测出烟雾,并做出早期的预警工作,且具有检测精确等优点,很方便在现实中推广使用。Smoke is the most prominent of the fire detect, so detect the smoke has an important significance in daily lives. In view of the current smoke detection has the high error rate and poor adaptability. Therefore, this paper presents Remove the fix interference which based on machine learning in the smoke detection and it can remove the fixed interference as soon as possible, and it can improve the correct rate of smoke recognition. Experiments show that the algorithm has a high speed and good effect on the smoke detection and make the early alarm work, And has accurate detection etc, so it＇ s convenient to be widely used in reality.

关 键 词：烟雾检测 运动块 线性分类器 固定干扰 机器学习 

分 类 号：TP274.2[自动化与计算机技术—检测技术与自动化装置]