室内环境下结合里程计的双目视觉SLAM研究

Research on Binocular Vision SLAM with Odometer in Indoor Environment

针对视觉SLAM要解决的定位精度低和鲁棒性低的问题,提出一种基于双目视觉传感器与里程计信息的扩展卡尔曼滤波SLAM方法,应用改进的SIFT算子提取双目视觉图像的环境特征获得特征点,并构建出视觉特征地图;应用扩展卡尔曼滤波算法融合视觉信息与机器人位姿信息,完成同时定位与地图创建。这种方法既可以解决单目视觉利用特殊初始化方法获取特征点信息不准确的问题,也可以避免双目视觉里程计利用图像信息恢复运动带来的计算量极大和运动估计不鲁棒的缺点。仿真实验表明,在未知室内环境下,算法运行稳定,定位精度高。

With the aim of solving the low positioning accuracy and low robustness problems of vision SLAM algorithm,Extended Kalman Filter（EKF） method based on binocular vision and odometer is proposed in this paper.Feature point can be obtained by extracting image features with improved SIFT algorithm,and the vision feature map is constituted.SLAM is completed by using the information of binocular vision and robot position with EKF.This method can either solve the monocular vision inaccuracy problem of feature point information obtained by special initialization method or avoid the enormous computation brought about by binocular vision odometer using image information to restore movement as well as the in-robust disadvantages of motion estimation.The results from simulation experiments indicate that in the unknown indoor environments,this algorithm operation is stable,and the positioning accuracy is high.