Simultaneous Localization and Mapping is a fundamental problem in mobile robotics. However, the majority of Visual SLAM algorithms assume a static scenario, limiting their applicability in real-world environments. Dealing with dynamic content in Visual SLAM is still an open problem, with solutions usually relying on purely geometric approaches. Deep learning techniques can improve the SLAM solution in environments with a priori dynamic objects, providing high-level information of the scene. However, most solutions are not prepared to deal with crowded scenarios. This paper presents Crowd-SLAM, a new approach to SLAM for crowded environments using object detection. The main objective is to achieve high accuracy while increasing the performance, in comparison with other methods. The system is built on ORB-SLAM2, a state-of-the-art SLAM system. The proposed methodology is evaluated using benchmark datasets, outperforming other Visual SLAM methods.

同时定位和映射是移动机器人技术中的一个基本问题。然而，大多数 Visual SLAM 算法都假设一个静态场景，限制了它们在现实环境中的适用性。在 Visual SLAM 中处理动态内容仍然是一个悬而未决的问题，解决方案通常依赖于纯几何方法。深度学习技术可以在具有先验的环境中改进 SLAM 解决方案动态对象，提供场景的高级信息。但是，大多数解决方案都不准备应对拥挤的情况。本文介绍了 Crowd-SLAM，这是一种使用对象检测针对拥挤环境进行 SLAM 的新方法。与其他方法相比，主要目标是在提高性能的同时实现高精度。该系统建立在最先进的 SLAM 系统 ORB-SLAM2 之上。所提出的方法是使用基准数据集进行评估的，其性能优于其他 Visual SLAM 方法。