@inproceedings{hhuang2020dualslam,
  author={Huang, Huajian and Lin, Wen-Yan and Liu, Siying and Zhang, Dong and Yeung, Sai-Kit},
  booktitle={2020 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS)}, 
  title={Dual-SLAM: A framework for robust single camera navigation}, 
  year={2020},
  volume={},
  number={},
  pages={4942-4949},
  doi={10.1109/IROS45743.2020.9341513}

}

This implementation is based on ORB-SLAM, an excellent feature-based monocular SLAM. The requirements are almost as same as ORB-SLAM. We have tested this implementation in Ubuntu 18.04.

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

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

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 3.2.

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

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.

GMS-Feature-Matcher-master

Clone the repository:

git clone https://github.com/HuajianUP/Dual_SLAM.git Dual_SLAM

Execute:

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

1. Download the dataset (grayscale images) from http://www.cvlibs.net/datasets/kitti/eval_odometry.php

2. In general, the more features are extracted, the more stable the system will be, but the processing time per frame will be longer. When extract 2000 features per frame, standard ORB-SLAM still potentially suffers from tracking loss and fail to constructs an intact map on several KITTI sequences (KITTI 00, 01, 02, 08, 09, 12, and 17). You could run the system on these sequences and see the recovery effect.

3. Execute the following command. Change KITTIX.yamlby KITTI00-02.yaml, KITTI03.yaml, KITTI04-12 or KITTI13-21.yaml for sequence 0 to 2, 3, 4 to 12, and 13 to 21 respectively.

4. In yaml files, change sequenceDir to your uncompressed sequence folder and imageDir to the image folder. Set recoveryFlag to 1 to activate recovery, initByGMS to 1 to change initialization algorithm.

./Examples/Dual-SLAM/dual_kitti Vocabulary/ORBvoc.bin config/KITTIX.yaml

You will need to create a settings file with the calibration of your camera. See the settings file provided for the KITTI datasets for monocular.

[ORB-SLAM] Raúl Mur-Artal, J. M. M. Montiel and Juan D. Tardós. ORB-SLAM: A Versatile and Accurate Monocular SLAM System. IEEE Transactions on Robotics, vol. 31, no. 5, pp. 1147-1163, 2015.PDF.