dydgns2017 / 3d-reconstruction-processing---intel-lidar-sensor

3d reconstruction use LiDAR sensor

3d-reconstruction-processing---intel-LiDAR-Sensor

3d reconstruction use LiDAR sensor

Hardware Information (datasheet)

Features
Use Environment	Technology	Ideal range
Indoor	Laser Scanning	.25m to 9m (Range affected by reflectivity)

Depth
Depth technology	Depth Field of View (FOV)	Minimum depth distance (Min-Z) at max resolution	Depth output resolution	Depth Accuracy	Depth frame rate
LiDAR	70° × 55° (±3°)	~25cm	Up to 1024*768	~5 mm to ~14 mm thru 9m**2	30fps

RGB
RGB frame resolution	RGB sensor FOV(H*V)	RGB frame rate	RGB sensor resolution	RGB sensor technology
1920 * 1080	70° × 43° (±3°)	30 fps	2 MP	Rolling Shutter

Major Components
Processing	Optical Board
Intel RealSense Vision ASIC	Laser & MEMS Mirror

Physical
Form Factor	Connectors	Diameter * Height	Mounting mechanism
Camera Peripheral	USB-C*3.1 Gen 1*	61mm * 26 mm	One 1/4‑20 UNC thread mounting point. & Two M3 thread mounting points. & Tripod

** reference : https://www.intelrealsense.com/lidar-camera-l515/

Process work

1. Hardware Settings
- Intrinsic Parameter
- LiDAR depth limit(?)
- Etc.
2. Export data (image & depth)
- export RGB(.jpg) & Depth (.png)
3. Make Fragments (.ply)
4. Make 3D-Object

Environment settings

• Python 3.8, open3d 0.15.1 with conda environment

conda create -n ro3d python=3.8
conda activate ro3d
pip install open3d
pip install joblib
pip install pyrealsense2
pip install opencv-python
pip install matplotlib

Open3d reconstruction example

• sample execution

git clone https://github.com/isl-org/Open3D
# open3d dir path set
cd open3d/
open3d_root=$(pwd)
cd examples/python/reconstruction_system
# config file path set
config_file=$open3d_root"/examples/python/reconstruction_system/config/tutorial.json"
# dataset download
# link : https://drive.google.com/file/d/11U8jEDYKvB5lXsK3L1rQcGTjp0YmRrzT/view
# you have to download dataset
# and then dataset will be unzip on datasets/016
# you should be change json file --> $config_file dataset path
# python run_system.py [config_file] [--make] [--register] [--refine] [--integrate]
run_system=$open3d_root"/examples/python/reconstruction_system/run_system.py"
cd $open3d_root
python $run_system $config_file --make
python $run_system $config_file --register
python $run_system $config_file --refine
python $run_system $config_file --integrate
python $run_system $config_file --slac
python $run_system $config_file --slac_integrate

• visualizing output result

cd $open3d_root && cd ..
python visualizer.py --filename datasets/016/slac/0.050/output_slac_pointcloud.ply
python visualizer.py --filename datasets/016/scene/integrated.ply

Create Own Dataset

• you should be ready dataset like below
• but, already setup in realsense_recorder.py file so you just use recorder script

datasets
└── realsense
    ├── camera_intrinsic.json
    ├── color
    │   ├── 000000.jpg
    │   ├── :
    └── depth
        ├── 000000.png
        ├── :

• recording dataset

cd $open3d_root && cd ..
python realsense_recorder.py --record_imgs

3D recon result from own dataset

• reconstruction from own dataset

# this is script setup config file for use recon system
# just copy or moify & paste terminal bash
cat << EOF > config.json
{
    "name": "Realsense image file",
    "path_dataset": "datasets/realsense",
    "path_intrinsic": "datasets/realsense/camera_intrinsic.json",
    "max_depth": 3.0,
    "voxel_size": 0.05,
    "max_depth_diff": 0.07,
    "preference_loop_closure_odometry": 0.1,
    "preference_loop_closure_registration": 5.0,
    "tsdf_cubic_size": 3.0,
    "icp_method": "color",
    "global_registration": "ransac",
    "python_multi_threading": true
}
EOF

• reconstruction by open3d script

cd open3d/ && open3d_root=$(pwd)
run_system=$open3d_root"/examples/python/reconstruction_system/run_system.py"
cd .. # go to github clone dir (HOME)

python $run_system config.json 
python $run_system config.json --make --register --refine --integrate

• result image

python visualizer.py --filename datasets/realsense/scene/integrated.ply

Reference

• Paper

- not yet update

• Blog Posting & Etc.

- https://www.intelrealsense.com/lidar-camera-l515/
- http://www.open3d.org/docs/release/tutorial/sensor/realsense.html
- https://darkpgmr.tistory.com/32 (Camera Calibration)
- http://www.open3d.org/docs/release/tutorial/reconstruction_system/system_overview.html#capture-your-own-dataset (open3d reconstruction)