Using Jing Wang's code as basis for VI-ORBSLAM2

1 August 2020

thesis

I’ve made some decisions in approach, listing them here with motivation:

I’m using Jing Wang’s VI-ORB implementation as a basis.

I will implement the VI methods for the stereo and RGB-D files. This means I will implement:

  cv::Mat System::TrackStereoVI(const cv::Mat &imLeft, const cv::Mat &imRight, const std::vector<IMUData> &vimu, const double &timestamp);
  cv::Mat System::TrackRGBDVI(const cv::Mat &im, const cv::Mat &depthmap, const std::vector<IMUData> &vimu, const double &timestamp);

  cv::Mat Tracking::GrabImageStereoVI(const cv::Mat &imLeft, const cv::Mat &imRight, const std::vector<IMUData> &vimu, const double &timestamp);
  cv::Mat Tracking::GrabImageRGBDVI(const cv::Mat &im, const cv::Mat &depthmap, const std::vector<IMUData> &vimu, const double &timestamp);

I will also make 2 ROS example executables, one for stereo and one for RGB-D cameras. This will include making the following:
- updating the Examples/ROS/ORB_VIO/CMakeLists.txt file
- ros_vio_stereo.cc and ros_vio_rgbd in Examples/ROS/ORB_VIO/src/
- renaming MsgSynchronizer to MonoMsgSynchronizer
- implementing StereoMsgSynchronizer to synchronize 2 images and a buffer of IMU messages