摘要
自然拍摄的视频信息通常包含前景和背景,采取传统轮廓提取或边缘提取等图像处理方法难以准确识别到视频信息中的人体轮廓信息。通过对比分析基于高斯混合模型法的OpenCV计算机视觉库中的BackgroundSubtractorKNN和BackgroundSubtractorMOG2背景减除算法,基于KNN算法提出了一种改进的背景减除算法,对所得图像经过中值滤波去除噪声影响后再进行阈值处理消除阴影部分从而提取出清晰的人体轮廓。并在此基础上基于像素比例系数关系提出了一种身高估算方法。实验表明,该方法能准确有效提取行走状态下的人体轮廓,身高估算与实际测量相对误差率在5%以内,能够较为准确的估算身高。
The natural video information usually includes foreground and background,and it is difficult to recognize the human body contour in the video information by traditional image processing methods such as contour extraction or edge extraction.In this paper,the background subtraction algorithms of BackgroundSubtractorKNN and BackgroundSubtractorMOG2 in OpenCV based on Gauss’ s hybrid model method are compared and analyzed,An improved background subtraction algorithm based on KNN algorithm is proposed.After the noise is removed by median filter,the shadow part is removed by threshold,and then a clear human body contour is extracted On this basis,a method of height estimation is proposed based on the relation of pixel scale coefficient.The experimental results show that the method can extract the human body contour accurately and effectively,and the relative error rate between height estimation and actual measurement is less than 5% height,Estimation can be more accurately.
作者
李忠浩
罗文田
陈乾
成鹏
LI Zhong-hao;LUO Wen-tian;CHEN Qian;CHENG Peng(Civil aviation flight university of China,electronic and electrical engineering institution,Guanghan 618307,China)
出处
《电脑与信息技术》
2022年第6期42-45,共4页
Computer and Information Technology
关键词
图像处理
混合高斯模型
背景消除
人体轮廓
身高估算
image processing
Gauss’s hybrid model
background subtraction
human body contour
height estimation
