项目名称: 基于单目多平面镜反折射成像系统的面向自动装配中复杂工件位姿计算
项目编号: No.61273283
项目类型: 面上项目
立项/批准年度: 2013
项目学科: 自动化技术、计算机技术
项目作者: 英向华
作者单位: 北京大学
项目金额: 83万元
中文摘要: 如何从工件箱中任意摆放的大量的不同种类工件里面,快速鲁棒地计算出待装配工件的三维空间位置和姿态,以利于机械手对其进行抓取,是工业自动装配领域亟待解决的一个最基本和最经典难题,即Bin-picking问题。随着机器视觉技术的发展,人们正朝着有效解决此难题的方向迈进。目前,由一台工业摄像机和多平面镜组成的反折射成像系统引起人们的关注。该系统可以在一帧图像中同时拍摄到工件的多个方位的像,既包括工件的实像也包括其多个镜中虚像。与传统的多摄像机系统相比,可以避免多摄像机标定难题以及同步问题,减少了系统成本。本课题将对反折射成像系统进行深入研究,包括:其内在成像机理;摄像机鲁棒标定;利用其特有的成像约束进行复杂物体检测、识别,以及精确的位姿计算等。通过以上内容的研究,建立一系列关于单目多平面镜反折射成像系统的有效算法,积极探索并试图解决目前工业自动装配领域中的经典难题,丰富机器视觉的基本理论和方法。
中文关键词: 三维重建;位姿估计;多视点几何;;
英文摘要: How to quickly and robustly calculate the 3D position and orientation of the pre-assembled workpiece in boxes for picking up, is one of the most fundamental and classic problems in the field of industrial automation assembly, namely, the Bin-picking problem. With the development of machine vision technology, peoples are seeking efficient solutions for the problem. A so-called catadioptric imaging system with an industrial camera and multiple plane mirrors, is very attractive. This imaging system can capture multiple views simultaneously in a single image, i.e., capture both the real and virtual views of the workpiece. Compared to traditional multiple camera system, the catadioptric imaging system can greatly reduce costs, avoid multiple camera calibration and synchronization problems. This proposal intends to investigate the catadioptric imaging system deeply, including: essential imaging mechanism; camera calibration; using unique imaging constraints for object detection, identification, and accurate pose calculation. We aim at establishing a series of efficient algorithms for the catadioptric imaging system, attempt to solve the classic problem in current industrial automatic assembly system, and enrich the essential theories and algorithms in machine vision.
英文关键词: 3D Reconstruction;Pose Estimation;Multiple View Geometry;;