RGB-D camera dense mapping based on ORB_SLAM2

chinese version

Preparation before starting

Confirm the build environment

Note: Whether you are using ubuntu18.04 or ubuntu16.04, it is strongly recommended to use the PCL 1.7.2 version of the point cloud library

• The actual measurement of pcl-1.8 installed by ubuntu18.04 apt will cause a segmentation fault (of course, it may also be the reason why my computer software environment is more complicated~)
• To compile and install pcl-1.7, refer to Compilation and installation steps on the pcl official website

Prepare the dataset

We use the RGB-D dataset of TUM: download it and put it under TUM_dataset/ in the project directory, if not, create a new one For details, please refer to the path of the command in the run_rgbd_dense_map.bash file.

Prepare dictionary file

If there is no dictionary file, then go to download: dictionary file

compile the entire project

     chmod +x build.sh
     ./build.sh

Compile only ORB_SLAM2 Dense_Map

    mkdir build
    cd build
    cmake..
    make -j

run rgbd dense map

    bash run_rgbd_dense_map.bash

running result:

The point cloud adjustment after the loopback is detected, the text output in the red box is the prompt information after the point cloud adjustment is completed after the loopback, see the code for details:

Code ideas

• Add a PointCloudMapping class and create a thread to process and display the point cloud information of keyframe
• The Tracking thread inserts keyframes into the PointCloudMapping keyframe queue
• The LoopClosing thread will notify PointCloudMapping to adjust the point cloud after a loop is detected.