基于视频处理技术的运动目标姿态检测与控制  

Posture detection and control of moving target based on video processing technology

作　　者：杨军[1] YANG Jun(Institute of Artificial Intelligence and Intelligent Rule of Law,Northwest University of Political Science and Law,Xi’an 710199,China)

机构地区：[1]西北政法大学人工智能与智慧法治研究院,西安710199

出　　处：《沈阳工业大学学报》2023年第1期90-96,共7页Journal of Shenyang University of Technology

基　　金：陕西省科技厅重点研发计划一般项目(2019GY-105)。

摘　　要：为了解决运动目标姿态检测常出现判定背景动态细节为运动目标的问题,提出了基于视频处理技术的运动目标姿态检测与控制方法.通过贝叶斯决策方法与帧间差分法相结合,划分视频图像为候选前景与候选背景,并设置该背景为主成分分析方法的初始值,获取目标姿态实时检测结果,依据检测结果设计PD控制器,进而通过控制参数实现目标检测和控制.结果表明,本文检测距离误差小于3 mm,角度误差小于1°,对比方法的误差分别为3~5 mm之间和3.2°~4.0°之间,高于本文方法.所提出的方法具有较高的检测精度,可实现运动目标姿态的有效控制.In order to solve the problem that the background dynamic details are often considered as moving objects in posture detection of moving objects,a method for posture detection and control of moving objects based on video processing technology was studied.In combination with the Bayesian decision method and the inter-frame difference method,the video image was divided into candidate foreground and candidate background,the initial value of background was set for the principal component analysis method,and the real-time detection result of target attitude was obtained.According to the detection result,the PD controller was designed,and then the target detection and control were realized through the parameter control.Experimental results show that the distance and angle errors of as-proposed method are less than 3 mm and 1°,respectively.The corresponding values reported by literatures are between 3 mm to 5 mm and 3.2°to 4.0°,respectively,higher than those of this work.Therefore,the as-proposed method has higher detection accuracy and can achieve the effective attitude control of moving target.

关 键 词：视频处理技术 运动目标 姿态实时检测 贝叶斯决策方法 帧间差分法 主成分分析方法 姿态控制 PD控制器 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]