论文部分内容阅读
针对复杂环境下运动机器人自主运行的任务要求,提出了一种采用图像特征匹配技术的视觉Sl AM算法。相比于传统滤波器方法在机器人长时间运动过程中产生的误差积累,采用了基于图优化的Sl AM方法,本文算法分为两个部分:前端和后端。前端负责处理图像数据提取机器人位姿几何关系,首先提取彩色RGB图像的特征点,创建特征点的高维特征描述子,建立特征点的位置对应关系;后端负责表达各时刻机器人的位姿并最大化地消除轨迹漂移问题,根据前端处理所产生的信息,构建一个代表机器人几何位置关系及其不确定性的关系图,通过对图的优化将测量轨迹最大化地逼近真实轨迹,最后生成稀疏点云地图和高精度的机器人运行轨迹。试验表明本文所提出的方法实用性强,具有较高的鲁棒性。
Aiming at the task requirements of autonomous autonomous robot in complex environment, a visual Sl AM algorithm using image feature matching is proposed. Compared with the traditional filter method in robot long-term movement caused by the accumulation of errors, the use of graph-based optimization of AM method, the algorithm is divided into two parts: the front and back. The front end is responsible for processing the geometric relationship between the robot and the robot. Firstly, the feature points of the RGB color image are extracted and the high-dimensional feature descriptors of the feature points are created to establish the position correspondence of the feature points. The rear end is responsible for expressing the pose of the robot at each moment According to the information generated by the front-end processing, a relational graph representing the geometric positional relationship of the robot and its uncertainty is constructed. By optimizing the graph, the measurement trajectory is maximally approximated to the real trajectory, and the sparse Point cloud map and high-precision robot trajectory. Experiments show that the method proposed in this paper is practical and has high robustness.