This project is Scan to BIM pipieline to convert 3D scan data (point cloud data) to BIM objects. The conventional process of converting 3D scan data into BIM models is often inefficient, requiring significant time and manual labor. This project was initiated to automate this process and maximize efficiency. I propose a flexible pipeline that utilizes a proprietary JSON-based script called SBDL (Scan to BIM Description Language). This allows users to dynamically define and execute processing steps, aiming to automate the entire workflow—from point cloud noise removal and object classification to geometry extraction and final BIM object generation. This was funded by KICT.
Scan to BIM research project has purpose like below.
In reference, you can use scan to model program (SMP), simple 3D point cloud to model pipeline version and use Lightweight 3D Point Cloud Segmentation Model PointEdgeSegNet for large-scale point cloud segmentation.
- 3D point cloud processsing pipeline implementation dynamically using simple SBDL(Scan to BIM Description Language. JSON format).
- Classification of outdoor building objects such as wall (facade), road etc.
- Extraction geometry information from classification.
- Binding BIM object with geometry information and property set.
v0.1
2022.11, Scan to BIM pipeline framework released. Simple SBDL was developed considering geometry computation algorithms to extract outdoor facade object, deep learning, docker based component etc.
v0.2
2023.7, Docker image support. pipeline revision for multiple input files processing. refactoring.
2023.8, Data augumentation tool
2023.9. LiDAR simulation tool
2023.10. 3D scan data quality checker tool.
2024.2. Update pcd_to_DTM, DTM_to_geo module to fix issue, add options such as "active", "log_view", "height_building_offset", "height_ground_offest", "max_building_height".
v0.3
documentation to use SBDL.
simple MLOps codes for outdoor object train.
v0.4
SBDL enhancement to supporting VFP(Visual Flow Programming) or LLM(Large Language Model. ex. ChatGPT).
Update indoor Object Mapping support.
MLOps support.
Simple Scan Data Processing App using Scan to BIM application.1) deep learing based indoor classification. 2) PCD indoor segmentation. 3) segment to geometry using ML. 4) geometry to BIM using revit plugin. 5) 3D data argumentation. 6) LiDAR simuation 7) 3D PCD quality check
v0.5
update PCD to DTM, DTM to Geometry, Geometry to BIM object source files.
- install python, pip
https://phoenixnap.com/kb/how-to-install-python-3-ubuntu - install cmake
https://www.cyberithub.com/how-to-install-cmake-on-ubuntu-20-04-lts-focal-fossa/ - install cuda, tensorflow, pytorch
- install gdal
https://mothergeo-py.readthedocs.io/en/latest/development/how-to/gdal-ubuntu-pkg.html - install pdal
https://installati.one/install-pdal-ubuntu-20-04/
In terminal, run 'pdal' command. If there is error 'libgdal.so.29: cannot open shared object file', run the below command to make linked file name.
sudo ln -s libgdal.so.30 libgdal.so.29
if there is error in prebuild pdal, download, build and install pdal source files of github https://github.com/PDAL/PDAL - install ifcopenshell
https://pypi.org/project/ifcopenshell/0.7.0.230418/
https://blenderbim.org/docs-python/ifcopenshell-python/installation.html - build docker image
cd docker
cd build_docker_open3d
bash build_docker.sh
In addition, if you use PCL, run the below commands for installing package or 'sh build_pcl.sh'.
sudo apt-get install build-essential g++ python3-dev autotools-dev libicu-dev libbz2-dev libboost-all-dev
sudo apt install libeigen3-dev
dpkg -L libeigen3-dev
sudo apt-get install -y libflann-dev
sudo apt-get install libpcap-dev
sudo apt-get install libgl1-mesa-dev
sudo apt install qtbase5-dev qtchooser qt5-qmake qtbase5-dev-tools
sudo apt install clang-format
sudo apt-get install libusb-1.0-0-dev
sudo apt install libvtk9.1
git clone https://github.com/PointCloudLibrary/pcl pcl-trunk
cd pcl-trunk && mkdir build && cd build
cmake -DCMAKE_BUILD_TYPE=RelWithDebInfo ..
make -j2
sudo make -j2 install
In detail, refer to
https://github.com/PointCloudLibrary/pcl
https://pcl.readthedocs.io/projects/tutorials/en/latest/compiling_pcl_posix.html
Modify the below PCL_ROOT(PCL library path), VTK_INC(VTK include path), USR_LIB paths in CMakeLists.txt properly. In reference, PCL-1.13 has an memory error (handmade_aligned_free) related to eigen library(2023/4/10).
set(PCL_ROOT "/home/ktw/projects/pcl-1.12")
set(VTK_INC "/usr/include/vtk-7.1")
set(USR_LIB "/usr/lib/x86_64-linux-gnu")
In terminal, input the below commands.
git clone https://github.com/mac999/scan_to_bim_pipeline
cd scan_to_bim_pipeline
pip -r install requirements.txt
sudo apt install clang
mkdir build
cmake ..
make
If there are depandency errors in requirements.txt, use requirements_simple.txt.
Before run, install requirements_simple.txt(or requirements.txt) including the above packages.
- modify /pipeline/config.json considering your input, output folder path. In reference, root foler name is scan_to_bim_pipeline which you downloaded and installed from github.
{
"app": "pcd_pipeline",
"root_path": "./pipeline/",
"bin_path": "./",
"lib_path": "./lib/",
"data_path": "./input/",
"debug_gui": false
}
- download input sample files and copy them into ./input folder. refer to sample dataset.
- run app.py like below.
python ./pipeline/app.py
- Or design pipeline by using SBDL(scan to bim description language) formatted by JSON like below.
pipeline.[name]={stage*}
stage={module_type, parameters}
parameters={name, value}*
module_type={python program | docker image | binary executable program}
- parameters should be defined in module before usage.
condition={"in_stage_return", "out_stage_return"}
In scan to BIM pipeline using SBDL example,
pipeline.indoor_obb_extraction = data_to_format > pcd_to_seg > pcd_to_clean > seg_to_geo
pipeline.indoor_obb_extraction(*.las) > *.geojson
{
"pipeline.indoor_obb_extraction": [
{
"type": "data_to_format",
"active": true,
"output_type": ".pcd"
},
{
"type": "pcd_to_seg",
"iteration": "1000",
"threshold": "0.1",
"projection": "true",
"remove_overlap_distance": "0.10",
"min_points_ratio": "0.2"
},
{
"type": "pcd_to_clean",
"voxel_down_size": "0.0",
"nb_radius_points": "50",
"nb_radius": "0.1"
},
{
"type":"seg_to_geo",
"alpha": "0.15"
}
]
}
cd pipeline
python app.py
Download dataset and copy to /input folder.
3D point cloud sample file download
National LiDAR map National Map
Open Topolography LiDAR map
Pix4D dataset Download
LAS map files arcgis map link
LiDAR files USGS gov
Top 6 Free LiDAR Data Sources LiDAR files
SBDL concept diagram and UML architecture.
MIT license.
Acknowledge.
Scan To BIM Technology Development 3D Urban Building Model Process Automation, 2022
3D vision & AI based Indoor object Scan to BIM pipeline for building facility management, 2023
Funded by KICT
Organization Roles
KICT: Scan to BIM pipeline architecture design, algorithm programming, test, code management
Specially, Thanks for contribution like below
IUPUI (Prof. Koo Dan, Prof. Kwonsik Song), UNF (Prof. Jonghoon Kim: usecase, code, policy survey
Purdue University (Prof. Kyubyung Kang): deep learning train, dataset collection, labeling, analysis
Stony Brook University (Prof. Jongsung Choi): data collection using SLAM, labeling, analysis
Kang, TW., Patil, S., Kang, K., Koo, D. and Kim, J., 2020. Rule-based scan-to-BIM mapping pipeline in the plumbing system. Applied Sciences, 10(21), p.7422. https://www.mdpi.com/2076-3417/10/21/7422
Kang, TW., 2023, Scan to BIM Mapping Process Description for Building Representation in 3D GIS, Applied Sciences. 13(17), https://www.mdpi.com/2076-3417/13/17/9986