These guidelines and procedures apply to lidar-camera calibration. Follow these guidelines for best results when using the Lidar and Camera Calibration workflow.
checkerboard function to create a
checkerboard image. The checkerboard must be a rectangle.
Use a checkerboard that contains an even number of squares along one edge and an odd number of squares along the other edge. A square pattern can produce unexpected results for camera extrinsics. A non-square pattern contains two black corners along one side and two white corners along the opposite side. Use these features to determine the orientation of the pattern and the origin point of the checkerboard. The calibrator assigns the x-direction to the longer side.
Print the checkerboard from end-to-end on a foam board to avoid any measurement errors.
Capture data from both sensors simultaneously. Capture data with no motion blur effects. Motion blur can degrade the accuracy of the calibration. If you are working with a video recording, carefully capture the point clouds corresponding to those frames.
Hold the checkerboard target with your arms fully extended, rather than close
to your body. Otherwise, parts of your body may appear to be planar with the
target. This could lead to inaccurate checkerboard detection. Hold the target
from behind, rather than on the edges, for sparse lidar sensors. The
detectRectangularPlanePoints function searches for planes in
each cluster. Specify the approximate checkerboard position with the ROI name-value
pair to reduce false detections.
Pay close attention to the distance between the sensor and the checkerboard. Distant checkerboards may not be detected accurately when you use low resolution lidar sensors like Velodyne® VLP-16.
Remove other items on the same plane that the calibrator might cluster with the target.
The figure below shows many different ways to hold the checkerboard for capturing data. Tilt the checkerboard to expose more of its area to the scan lines from the lidar sensor. Capture at least 10 frames for a proper calibration.
Be aware of the viewing angle or the field of view of the lidar. Make sure that the boards are not placed in the blindspots of the sensor.
For high-resolution lidar sensors like the HDL-64 and Ouster OS1-64, you can hold the checkerboard horizontally or vertically while capturing data. However, tilt the checkerboard to a 45 degree angle while capturing data for best results.
Point cloud data must be saved in the PCD or PLY format.
If the data was captured in rosbag format, you can use the Read Lidar and Camera Data from Rosbag File tutorial to load calibration data into the app.