摘要
研究基于改进GMM的智能变电站视频目标检测方法,该方法精准检测变电站视频目标。构建智能变电站视频序列各像素的高斯分布模型,通过判断像素点与高斯分布模型匹配情况,确定该像素点是否为背景,利用四邻域法调整学习率,实现首个匹配高斯分布模型权值调整,根据权值与方差的比值大小实现背景高斯分布模型的降序排列,获取智能变电站视频背景图像,当前帧图像除去背景图像即可得到前景图像,通过去除前景图像的离散点噪声、视频目标内部噪声和阴影,检测智能变电站视频目标。实验结果表明,像素高斯分布为5时,该方法检测到的智能变电站前景图像中的视频目标效果更好,错检像素、漏检前景像素平均值均较低。
The video target detection method of intelligent substation is studied based on improved GMM to accurately detect the video target of substation.A Gaussian distribution model is built for each pixel of the video sequence of the intelligent substation,and to determine whether the pixel is the background by judging the matching between the pixel and the Gaussian distribution model.It can adjust the learning rate by the four neighborhood method,realize the weight adjustment of the first matching Gaussian distribution model,and realize the descending arrangement of the background Gaussian distribution model according to the ratio of weight value to variance.It can also obtain the video background image of the intelligent substation,remove the background image from the current frame image to obtain the foreground image,and detect the video target of the intelligent substation by removing the discrete point noise,internal noise and shadow of the foreground image.The experimental results show that when the Gaussian distribution of pixels is 5,the video target effect in the foreground image of intelligent substation detected by this method is better,and the average values of false detection pixels and missed detection foreground pixels are lower.
作者
李聪
LI Cong(State Grid Beijing Electric Power Company,Beijing 100051,China)
出处
《微型电脑应用》
2024年第5期88-91,共4页
Microcomputer Applications
基金
国网北京市电力公司科技项目计划任务书(52021621003Q)。
关键词
改进GMM
智能变电站
视频目标检测
高斯分布模型
improved GMM
intelligent substation
video target detection
Gaussian distribution model
