计算机工程与应用2024，Vol.60Issue(15)：318-326,9.DOI:10.3778/j.issn.1002-8331.2305-0080
感兴趣区域与距离图像融合的点云基准标记检测系统
Pointcloud Fiducial Marker Detection System for Fusing Region of Interest with Range Image
摘要
Abstract
Traditional visual fiducial marker systems use cameras as sensors,which are susceptible to the impact of scene lighting and image quality.The pointcloud fiducial marker system uses 3D LiDAR as the sensor,which has higher stability and robustness,but so far it is rare.LIDARTag has spatial limitations on the placement of fiducial marker.The IILFM system eliminates the limitations on the LIDARTag,but using intensity gradient to extract features has a large number of outliers,which affects marker detection.Additionally,using ICP algorithm for pose estimation has significant errors in datasets containing dynamic point clouds.In order to eliminate the limitations of LIDARTag and achieve an efficient,robust,and flexible fiducial marker system,this paper improves the IILFM by using the intensity threshold of fiducial markers,as well as the local feature descriptor of point cloud PFH,as constraints to segment regions of interest.The mapping between 3D point cloud and 2D range image is achieved through spherical projection.Afterwards,the 6D pose estimation method based on relative geometric corner invariance is expanded to be suitable for solving the pose estimation problem.The qualitative and quantitative experimental results indicate that this system can eliminate the limitations of LIDARTag.Compared to the IILFM system,this method can eliminate outliers in the features.Compared to the visual reference mark detection system,this system is not affected by lighting and has more robust pose estimation accuracy.关键词
点云基准标记检测系统/强度梯度/PFH描述子/感兴趣区域/距离图像/六自由度位姿估计Key words
pointcloud fiducial marker system/intensity threshold/PFH descriptor/region of interest/range image/6D pose estimation分类
信息技术与安全科学引用本文复制引用
刘博文,曾碧,刘建圻..感兴趣区域与距离图像融合的点云基准标记检测系统[J].计算机工程与应用,2024,60(15):318-326,9.基金项目
国家自然科学基金(62172111) （62172111）
中山市重大科技专项(191018182628219). （191018182628219）
