【摘要】 研究了一种基于松组合的视觉惯性即时定位与同步构图(SLAM)方法。针对视觉特征点匹配率低问题,研究基于ORB(Oriented FAST and Rotated BRIEF)特征点的提取方法;基于ORB-SLAM的输出,结合SINS提出了一种具有回环检测功能的SLAM/SINS组合方法。利用最小二乘法估计视觉SLAM算法的尺度因子;构建SLAM/SINS的非线性卡尔曼滤波器,将视觉SLAM系统输出的位置信息经过尺度变换后作为观测量进行卡尔曼滤波,修正惯导的误差。最后利用标准数据集证明与开源的SLAM算法进行对比,结果表明,所提出的算法有比较高的定位精度,并且在移动设备上开发了增强现实软件,以增强现实为实验手段验证在较大的空间范围和环境干扰下,这种组合方法具备较好的漂移消除能力。更多还原

【Abstract】 A visual-aid inertial SLAM method based on loose couple is presented. Due to the low matching rate of visual feature points, the feature extraction method based on ORB(Oriented FAST and Rotated BRIEF) is studied and SLAM/SINS combinational navigation with loop detection function based on ORB-SLAM outputs and SINS is proposed. The scale factor is estimated by least squares and a nonlinear Kalman filter of SLAM/SINS is constructed. The output of ORB-SLAM is transformed into the observation of Kalman filter and the error of SINS is corrected. The experiment uses standard dataset to prove its high accuracy and also tests the algorithm by developing AR software in the phone. The results show that the proposed three dimension registration technology has better real-time and robustness under the large spatial range and environmental interference.更多还原