FAST Dynamic ORB-SLAM2

Authors: Justin Bi, Li Chen, Pradeep Suresh, Yicheng Tao, Yuliang Zhu

FAST Dynamic ORB-SLAM2 is a near real-time SLAM library that is built on ORB-SLAM 2 by Mur-Artal et al. It attempts to mitigate the error introduced by dynamic objects in the SLAM system on RGB-D cameras. See our other repository for related work: https://github.com/bijustin/YOLO-DynaSLAM

Our paper is located here: https://github.com/bijustin/Fast-Dynamic-ORB-SLAM/blob/master/dynamic-orb-slam.pdf

Our video is located here: https://www.youtube.com/watch?v=sazkFVVjITQ

Moving Object Detection

In the project, Firstly, ORB featurematches between two close frames are found. Secondly,the fundamental matrix is solved using the RANSACalgorithm. The third step is calculating the epipolar lines(corresponding to the outlier matches) in the secondframe, and calculate the distance between the feature points and the epipolar lines. If the distance is greaterthan some threshold, then we consider its corresponding feature point to be part of the moving object.

Finally, we utilize optical flow's angle's relation to optimize our results.

The following are the results using TUM dataset.

1. License

FAST Dynamic ORB-SLAM2 is based on ORB-SLAM2, and as such is released under a GPLv3 license. For a list of all code/library dependencies (and associated licenses), please see Dependencies.md.

2. Prerequisites

FAST Dynamic ORB-SLAM 2 has been tested in Ubuntu 18.04, but should be fine to compile on other platforms. A powerful CPU will result in real-time results.

Libraries

FAST Dynamic ORB-SLAM 2 is developed based on ORB-SLAM 2, and uses no additional libraries. Thus, the required libraries are the same, as follows:

C++11 or C++0x Compiler

We use the new thread and chrono functionalities of C++11.

Pangolin

We use Pangolin for visualization and user interface. Dowload and install instructions can be found at: https://github.com/stevenlovegrove/Pangolin.

OpenCV

We use OpenCV to manipulate images and features. Dowload and install instructions can be found at: http://opencv.org. Required at leat 2.4.3. Tested with OpenCV 2.4.11 and OpenCV 3.2.

Eigen3

Required by g2o (see below). Download and install instructions can be found at: http://eigen.tuxfamily.org. Required at least 3.1.0.

DBoW2 and g2o (Included in Thirdparty folder)

We use modified versions of the DBoW2 library to perform place recognition and g2o library to perform non-linear optimizations. Both modified libraries (which are BSD) are included in the Thirdparty folder.

3. Building FAST Dynamic ORB-SLAM2 library and examples

Clone the repository:

git clone https://github.com/bijustin/Fast-Dynamic-ORB-SLAM.git FASTORB_SLAM2

We provide a script build.sh to build the Thirdparty libraries and FAST Dynamic ORB-SLAM2. Please make sure you have installed all required dependencies (see section 2). Execute:

cd FASTORB_SLAM2
chmod +x build.sh
./build.sh

This will create libORB_SLAM2.so at lib folder and the executable rgbd_tum in Examples folder.

6. RGB-D Example

TUM Dataset

1. Download a sequence from http://vision.in.tum.de/data/datasets/rgbd-dataset/download and uncompress it.

2. Associate RGB images and depth images using the python script associate.py. We already provide associations for some of the sequences in Examples/RGB-D/associations/. You can generate your own associations file executing:

python associate.py PATH_TO_SEQUENCE/rgb.txt PATH_TO_SEQUENCE/depth.txt > associations.txt

3. Execute the following command. Change TUMX.yaml to TUM1.yaml,TUM2.yaml or TUM3.yaml for freiburg1, freiburg2 and freiburg3 sequences respectively. Change PATH_TO_SEQUENCE_FOLDERto the uncompressed sequence folder. Change ASSOCIATIONS_FILE to the path to the corresponding associations file.

./Examples/RGB-D/rgbd_tum Vocabulary/ORBvoc.txt Examples/RGB-D/TUMX.yaml PATH_TO_SEQUENCE_FOLDER ASSOCIATIONS_FILE

Acknowledgements

Our code builds on ORB-SLAM2 and DynaSLAM.