Dynamic Object Tracking and Masking for Visual SLAM

TL;DR

This paper introduces a fast, neural network-based pipeline that enhances visual SLAM in dynamic environments by removing moving objects from the map, improving localization, mapping, and dynamic object tracking.

Contribution

It presents a novel, efficient method combining deep learning and filtering techniques to improve SLAM performance in dynamic scenes, including dynamic object masking and 3D mapping.

Findings

Achieves 14 fps on a GTX 1080

Maintains state-of-the-art localization accuracy

Produces dynamic-object-free 3D maps

Abstract

In dynamic environments, performance of visual SLAM techniques can be impaired by visual features taken from moving objects. One solution is to identify those objects so that their visual features can be removed for localization and mapping. This paper presents a simple and fast pipeline that uses deep neural networks, extended Kalman filters and visual SLAM to improve both localization and mapping in dynamic environments (around 14 fps on a GTX 1080). Results on the dynamic sequences from the TUM dataset using RTAB-Map as visual SLAM suggest that the approach achieves similar localization performance compared to other state-of-the-art methods, while also providing the position of the tracked dynamic objects, a 3D map free of those dynamic objects, better loop closure detection with the whole pipeline able to run on a robot moving at moderate speed.