摘要
由于柔性物体在不同时刻、不同区域的运动状态是不同的,给计算机视觉研究带来了困难。针对这一问题,本文提出一种将改进的LK光流法与双目视觉原理相结合的方法研究面料在空间中的三维运动。首先,在时间序列重建中采用改进LK光流法将输入的图像高通滤波,滤除图像低频的照射分量,再将两帧之间使用加权最小二乘法获取光流,计算目标物体三维运动中的速度、加速度;其次,在空间重建中,利用三角测量法计算二维图像的三维信息,实现点云重构。实验结果表明,该方法提高了匹配精度与效率,通过对面料的重建,获取目标精确的运动信息,进而全面准确的分析其运动情况。
Due to the different motion states of flexible objects at different times and in different regions,it brings difficulty to the study of computer vision at present.In this paper,an improved LK optical flowmethod combined with the principle of binocular vision is proposed to study the three-dimensional movement of fabric in space.Firstly,in the reconstruction of time series,the input image is filtered by the improved LK optical flowmethod to remove the low-frequency illumination component of the image,then the weighted least square method is used to obtain the optical flowbetween the two frames to calculate the speed and acceleration of the three-dimensional movement of the target object.Secondly,in the spatial reconstruction,the three-dimensional information of the two-dimensional image is calculated by the triangulation method to realize the point cloud reconstruction.The experimental results showthat this method improves the matching accuracy and efficiency.Through the reconstruction of the fabric,the accurate motion information of the target can be obtained,and then the motion of the target can be analyzed comprehensively and accurately.
作者
薛冕
刘翔
XUE Mian;LIU Xiang(School of Electrical and Electronic Engineering,Shanghai University of Engineering Science,Shanghai 201620,China)
出处
《智能计算机与应用》
2020年第7期56-60,共5页
Intelligent Computer and Applications
基金
国家自然科学基金(61876106)
上海市科研计划项目(19ZR1421500)
关键词
三角测量
光流法
运动
加权最小二乘法
Triangulation principle
Optical flow
Motion
Weighted least squares
