Real-time SLAM Pipeline in Dynamics Environment

TL;DR

This paper introduces a real-time SLAM pipeline capable of operating in dynamic environments by integrating RGB-D SLAM with YOLO object detection to segment and remove moving objects, enabling accurate static scene reconstruction.

Contribution

It presents a novel real-time SLAM method that combines semantic object detection with geometric filtering to handle dynamic scenes, which was not addressed in prior static scene SLAM approaches.

Findings

Successfully reconstructs static scenes in dynamic environments

Demonstrates real-time performance with integrated YOLO and RGB-D SLAM

Provides a new dataset for semantic, geometric, and physical scene analysis

Abstract

Inspired by the recent success of application of dense data approach by using ORB-SLAM and RGB-D SLAM, we propose a better pipeline of real-time SLAM in dynamics environment. Different from previous SLAM which can only handle static scenes, we are presenting a solution which use RGB-D SLAM as well as YOLO real-time object detection to segment and remove dynamic scene and then construct static scene 3D. We gathered a dataset which allows us to jointly consider semantics, geometry, and physics and thus enables us to reconstruct the static scene while filtering out all dynamic objects.