Go2Py_SIM/docs/utlidar-camera-calib.md

57 lines
1.8 KiB
Markdown

# Lidar to Camera Extrinsic Calibration
The goal of this procedure is to find the extrinsic transformation between the Lidar coordinate frame and that of the camera in front of the robot:
<p align="center">
<img src="assets/lidar-camera-extrinsics.drawio.png" alt="image" width="60%" height="auto"/>
</p>
First pull the docker image:
```bash
docker pull koide3/direct_visual_lidar_calibration:humble
```
Then activate the X server connection port:
```bash
xhost +
```
and run the container with GUI support:
```bash
docker run \
-it \
--rm \
--net host \
--gpus all \
-e DISPLAY=$DISPLAY \
-v $HOME/.Xauthority:/root/.Xauthority \
-v /path/to/input/bags:/tmp/input_bags \
-v /path/to/save/result:/tmp/preprocessed \
koide3/direct_visual_lidar_calibration:humble
```
Inside the container generate images from the LiDAR data (Preprocessing):
```bash
ros2 run direct_visual_lidar_calibration preprocess /path/to/bag/dir /path/to/result/dir \
--image_topic /camera/color/image_raw \
--points_topic /utlidar/cloud \
--camera_model plumb_bob \
--camera_intrinsics 379.7099304199219,320.3064270019531,379.3695983886719,243.11753845214844 \
--camera_distortion_coeffs -0.057967256754636765,0.0704321563243866,-0.00015285948757082224,0.0006057045538909733,-0.022366832941770554
```
then manually select the correspondences between the 3D pointcloud and the image:
```bash
ros2 run direct_visual_lidar_calibration initial_guess_manual /path/to/result/dir
```
Finally perform a refinement step by running:
```bash
ros2 run direct_visual_lidar_calibration calibrate /path/to/result/dir
```
[![Alt text](https://img.youtube.com/vi/YOUTUBE_VIDEO_ID/0.jpg)](https://www.youtube.com/watch?v=FTlC9RwEVxY&t=43s)
# Source
This procedure has been taken from [direct_visual_lidar_calibration toolbox](https://koide3.github.io/direct_visual_lidar_calibration/).