摘要
为提升运动动作检测及跟踪效果,文中提出基于高斯混合模型的运动动作跟踪方法,该方法利用高斯混合模型经参数初始化、参数更新、背景选取以及前景检测等步骤,确定运动动作目标前景区域,获得清晰度较高的图像运动动作检测结果;并采用卡尔曼滤波算法,经运动目标外接矩形标定、特征信息计算提取运动特征后,通过构建帧间关系矩阵,预测估计运动动作区域,实现运动动作跟踪并输出结果。经实验验证,该方法在图像预处理阶段能够获得较为清晰的待跟踪图形,且运动动作检测准确率较高,跟踪误差小,运动动作跟踪效果好。
In order to improve the effect of motion detection and tracking,a motion tracking method based on Gaussian mixture model is proposed in this paper.The method uses Gaussian mixture model to determine the front scenic area of motion object through parameter initialization,parameter update,background selection and foreground detection,so as to obtain high-definition image motion detection results;The Kalman filter algorithm is used to extract the motion features through the calibration of the moving object’s external rectangle and the calculation of the feature information.Then the motion action region is predicted and estimated by constructing the inter frame relation matrix to realize the motion tracking and output the results.The experiment results show that this method can get a clear image to be tracked in the image preprocessing stage,and the motion detection accuracy is high,the tracking error is small,and the motion tracking effect is good.
作者
杜保亮
DU Bao-liang(China University of Petroleum,Qingdao 266555,Shandong Province,China)
出处
《信息技术》
2022年第7期1-5,11,共6页
Information Technology
基金
山东省重点研发计划(软科学项目)(2019RKE28099)。
