基于几何约束的室内动态环境视觉SLAM

Geometric Constraint-Based Visual SLAM Under Dynamic Indoor Environment

同时定位与地图创建(Simultaneous Localization and Mapping,SLAM)作为自主移动机器人的基本功能,近年来已成为机器人领域的研究热点。然而现有视觉SLAM算法大多将外部场景作为静态假设,忽略了环境中运动物体对SLAM系统精度的影响,影响SLAM系统在实际环境中的应用。鉴于此,提出一种适用于动态场景的鲁棒视觉SLAM算法,以ORB-SLAM2框架RGB-D模式为基础,在前端添加一种基于几何约束的动态检测方法。首先对场景中的动态特征点利用一种几何约束方法进行粗滤除,然后将剩余的特征点作为改进随机抽样一致算法(Random Sample Consensus,RANSAC)的样本点估算稳定的基本矩阵,使用极线几何滤除场景中真正的动态特征点。最后对改进系统在TUM数据集上进行测试,结果表明改进系统可以有效提高ORB-SLAM2系统在高动态场景中的性能。

As a basic function of autonomous mobile robots,Simultaneous Localization and Mapping(SLAM)has been widely researched in recent years.However,most state-of-art visual SLAMs adopt a strong scene rigidity assumption for analytical convenience,which limits the utility of these algorithms for real-world environments with independent dynamic objects.This paper presents a robust visual SLAM towards dynamic indoor scenes,which is built on the RGB-D mode of ORB-SLAM2.A dynamic detection method based on geometric constraints is added to the front end of ORB-SLAM2.First,the dynamic features in the scene are coarsely filtered using a geometric constraint method.Then the remaining features are used as sample points for the improved Random Sample Consensus(RANSAC)algorithm to estimate the stable fundamental matrix.And the epipolar geometry is used to filter out the real dynamic features in the scene.Experiments on the public TUM RGB-D dataset are conducted to evaluate the proposed approach.This evaluation reveals that the proposed algorithm can effectively improve the positioning accuracy of the ORB-SLAM2 system in high-dynamic scenarios.