摘要
快速准确地判断与目标平面的方位角是移动机器人执行后续任务的关键。现有的方位角测量方法对特征提取的质量有很高的要求,过于依赖特征提取的质量不仅使检测速度受到限制,也会由于特征的缺失增加检测的误差。通过对视觉平面测量方法研究,提出通过拍摄一张粘贴在目标平面上的已知尺寸矩形靶标图像来快速解算相机平面和目标平面间的夹角。首先获取矩形靶标上的特征点,根据针孔成像原理,解算其到相机光心的距离,建立相机平面和矩形平面在三维空间中的数学模型,求取相机平面和矩形平面间的偏转角度。通过实验验证,相机与矩形靶标分别相距1 m、2 m和3 m时,在0~60°偏转范围内的最大偏转角误差为0.984 0°,程序耗时在30 ms以内,能满足多数定位的快速准确性要求。
The key point for mobile robots to perform subsequent tasks is determining the azimuth between the robot and the target plane quickly and accurately.Existing visual angle deflection measurement methods have high requirements on the quality of feature extraction.However,over-reliance on the quality of feature extraction not only limits the detection speed but also increases the probability of mislocalization due to missing features.Through studying current planimetric methods,a method is proposed to quickly solve the angle between the camera plane and the target plane by taking one image of a rectangular target of known size pasted on the target plane.The mathematical model of the camera plane and the rectangular plane in 3D space is established by obtaining the feature points on the rectangular target and solving the distance from them to the camera optical center,and then the deflection angle between the camera plane and the rectangular plane is found according to the pinhole imaging principle.Through experiments,the maximum deflection angle error in the range of 0~60°deflection is 0.9840°when the camera and the rectangular target are 1 m,2 m and 3 m apart respectively,and the procedure takes less than 30 ms,which can meet most of the requirements for fast and accurate positioning.
作者
吴金坤
李忠国
席茜
车赛
石周
WU Jinkun;LI Zhongguo;XI Qian;CHE Sai;SHI Zhou(School of Mechanical Engineering,Jiangsu University of Science and Technology,Zhenjiang 212100)
出处
《计算机与数字工程》
2025年第6期1697-1703,共7页
Computer & Digital Engineering
基金
国家重点研发计划项目(编号:2018YFC0309100)资助。
关键词
单目视觉
矩形靶标
图像处理
角度测量
monocular vision
rectangular targets
image processing
angle measurement
作者简介
吴金坤,男,硕士研究生,研究方向:数字图像处理;李忠国,男,博士,副教授,研究方向:基于图像的目标跟踪和识别、视觉导引;席茜,女,硕士研究生,研究方向:数字图像处理;车赛,男,硕士研究生,研究方向:水下遮挡目标的分割与修补;石周,男,硕士研究生,研究方向:数字图像处理。
