摘要
为了实现对任意摆放的物体的识别与位姿估计,需要得到物体的三维模型,提出了一种高效准确的物体三维点云模型构建方法。首先通过RGB-D传感器获取物体多角度数据,然后利用尺度不变特征变换(SIFT)特征点匹配和改进的迭代最近点(ICP)算法计算出各角度下传感器的相对位姿,进而生成目标物体所在场景的完整点云,并通过物体分割和点云后处理得到目标物体的三维点云模型。实验结果显示,构建的物体点云模型清晰且不失真,并保留了表面完整的特征信息。
To realize the recognition and pose estimation of arbitrary objects, a three-dimensional model of the object is needed, and an efficient and accurate 3D point cloud model building method is proposed. First, the multi view data of the object is obtained by the RGB-D sensor, and the relative pose of the sensor at various angles are calculated by using scale invariant feature transform (SIFT) feature point matching and the improved iterative closest point (ICP) algorithm to generate the complete point cloud of the scene where the target object is located, and the 3D point cloud model of the object is further obtained by the object segmentation and the point cloud after processing. The experiment results show that the point cloud model is clear and undistorted, retaining the complete feature information of the surface of the object.
作者
于灏
杜华军
蔡莹皓
鲁涛
王睿
王硕
Yu Hao;Du Huajun;Cai Yinghao;Lu Tao;Wang Rui;Wang Shuo(State Key Laboratory of Management and Control for Complex System, Institute of Automation,Chinese Academy of Sciences, Beijing 100190;University of Chinese Academy of Sciences, Beijing 100049;Center for Excellence in Brain Science and Intelligence Technology,Chinese Academy of Sciences, Shanghai 200031;Beijing Aerospace Automatic Control Institute, Beijing 100854;National Key Laboratory of Science and Technology on Aerospace Intelligent Control, Beijing 100854)
出处
《高技术通讯》
EI
CAS
北大核心
2019年第8期750-757,共8页
Chinese High Technology Letters
基金
国家自然科学基金(U1713222,61773378)资助项目
