摘要
GMM算法(高斯混合模型算法)是一种用于背景建模的高效算法,然而传统的GMM算法比较适合于背景很少发生变化的情况,由于无人机这种高速移动平台自身的特殊性,背景时时刻刻在发生变化,因而导致GMM算法会出现很多的误判,无法适应这种复杂多变的环境。为了解决这个问题,提出一种基于GMM的多颜色空间融合的火灾检测算法。首先使用HSV、XYZ等多种颜色模型和形态学方法对无人机拍摄的视频帧图像进行预处理,然后在此基础上使用改进的三帧差分法配合改进的自适应GMM算法进行烟雾和火焰的检测,最后使用形态学方法进一步去除噪声。与传统的GMM算法相比,该算法能够有效地满足无人机高速移动平台对于算法实时性和检测性能的要求,能够很好地去除噪声,快速、准确地检测出移动的烟雾和火焰。
The GMM algorithm(Gaussian Mixture Model)is an efficient algorithm for background modeling,but the traditional GMM algorithm is more suitable for the background rarely changes.Due to the particularity of the high-speed mobile platform of the UAV,the background is constantly changing,which leads to a lot of misjudgments in the GMM,unable to adapt to this complex and changeable environment.In order to solve this problem,we propose a fire detection algorithm based on GMM and multi-color space fusion.Firstly,the color models like HSV,XYZ and morphological methods are used to preprocess the video frame images taken by the drone,and then an improved three-frame difference method with an improved adaptive GMM algorithm are applied to detect smoke and flame.Finally,morphological method is to further remove noise.Compared with the traditional GMM algorithm,the proposed algorithm can effectively meet the real-time and detection performance requirements of the UAV high-speed mobile platform,remove noise well,and detect moving smoke and flame quickly and accurately.
作者
金仙力
宋少杰
刘林峰
JIN Xian-li;SONG Shao-jie;LIU Lin-feng(School of Computer,Nanjing University of Posts and Telecommunications,Nanjing 210023,China)
出处
《计算机技术与发展》
2022年第7期75-81,共7页
Computer Technology and Development
基金
国家自然科学基金面上项目(61872191)。
关键词
颜色模型
三帧差分法
自适应高斯混合模型
形态学方法
运动检测
无人机
color model
three-frame difference method
adaptive Gaussian mixture model
morphological method
motion detection
UAV
