基于机器视觉技术的复杂运动姿态自动跟踪研究  

Complex motion attitude automatic tracking based on machine vision technology

作　　者：张炜 陈晨[2] ZHANG Wei;CHEN Chen(Inner Mongolia Normal University,Hohhot 010051,China;Baotou Railway Vocational&Technical College,Baotou 014060,China)

机构地区：[1]内蒙古师范大学,内蒙古呼和浩特010051 [2]包头铁道职业技术学院,内蒙古包头014060

出　　处：《现代电子技术》2021年第11期62-66,共5页Modern Electronics Technique

摘　　要：以提高复杂运动姿态自动跟踪精度、避免遮挡和信息丢失问题为出发点,研究基于机器视觉技术实现的运动姿态自动跟踪方法。借助摄像机采集运动图像,以机器视觉技术具备较好鲁棒性的HSV色彩空间为基础,采用中值滤波对图片实施预处理,有效去除噪声干扰;利用三帧差法检测预处理完成图像间的灰度差,以此确定图像中运动目标位置后,利用退火粒子滤波算法实现对复杂运动姿态的自动跟踪。实验结果表明,该方法可有效去除所采集图像中的噪声干扰,保留图像细节信息,且运动图像检测均方误差较小,粒子数和退火层数分别为100个和5层时,该方法的跟踪误差最小,约束条件的有效加入,可以解决运动过程中出现的自遮挡、深度信息丢失等问题。A motion attitude automatic tracking method based on machine vision technology is studied to improve the automatic tracking accuracy of complex motion attitude and avoid occlusion and information loss.In the method,the motion images are collected with a video camera.On the basis of the machine vision technology and based on HSV color space with good robustness,the median filtering is used to preprocess the images to effectively remove the noise interference.The three frame difference method is used to detect the gray difference between preprocessed images to determine the position of moving objects in the image,and then the annealed particle filtering algorithm is used to automatically track the complex motion attitude.The experimental results show that the proposed method can effectively eliminate the noise interference in the collected images and retain the image details;its mean square error of the motion image detection is small;its tracking error is the minimum when the number of particles and annealing layers are 100 and 5,respectively;the self-occlusion and depth information loss appeared in the process of motion can be avoided by adding constraint conditions into the proposed algorithm effectively.

关 键 词：自动跟踪 复杂运动姿态 机器视觉 运动图像检测 降噪预处理 退火粒子滤波 

分 类 号：TN911.73-34[电子电信—通信与信息系统]