摘要
即时定位与地图构建(SLAM)是解决移动机器人在未知非结构化环境中自主导航与控制的关键,一个完整的SLAM系统包括传感器数据处理、位姿估计、构建地图、回环检测四个部分。其中回环检测机制是解决移动机器人的闭环重定位,提高SLAM系统鲁棒性的重要环节。该研究提出一种基于ORB词袋模型的SLAM系统框架,通过研究与分析了使用FLANN算法选取关键帧与匹配帧间特征点,ORB特征描述子对检测速度的提高,通过k-means++算法对特征点进行训练生成含有视觉单词的词袋模型,使用高斯金字塔的直方图交叉核的SVM分类器,使用e PNP算法的增量式帧间位姿估计,回环检测重定位机制等环节,实现了单目视觉SLAM系统的初始化与位姿优化,实现了在丢帧状况下通过词袋模型进行重定位。最后通过搭建实验平台和标准数据集的测试得到的数据结果表明,基于ORB词袋模型的SLAM系统,具有良好的实时性,能够有效提高SLAM系统的重定位准确性,增强了系统的鲁棒性。
SimuItaneous localization and mapping （SLAM） is the key to solve the autonomous navigation and control of mobile robots in an unknown unstructured environment, A complete SLAM system includes four parts： sensor data processing, pose es-timation, mapping, loop-closure. The loop-closure detection mechanism is an important part o f solving the closed-loop reloca-tion of mobile robots and improving the robustness o f SLAM system. This study proposes a SLAM system framework based on ORB bag of words model, Select the keyframe and match the interframe feature points by studying and analyzing FLANN al-gorithm, The ORB characterization enhances the detection speed, Train the feature points to generate the word pocket model with visual words by k-means＋ ＋ algorithm, The SVM classifier using Histogram intersection kernel with Gaussian pyramid, In-cremental Interframe Pose Estimation using ePNP Algorithm, relocation by closure-loop detection mechanism, achieve a mono-cular vision SLAM system initialization and pose optimization, And the relocation is earned out by the word pocket model under the frame dropping situation. Finally, through the construction o f the experimental platform and the standard data set test results show that, SLAM System Based on ORB bag o f words,has a good real- time,can effectively improve the SLAM system relo-cation accuracy, and enhance the robustness o f the system.
出处
《信息通信》
2017年第10期20-25,共6页
Information & Communications
