search

York-SDCNLab / IILFM

This is a fiducial marker system designed for LiDAR sensors. Different visual fiducial marker systems (Apriltag, ArUco, CCTag, etc.) can be easily embedded. The usage is as convenient as that of the visual fiducial marker. The system shows potential in SLAM, multi-sensor calibration, augmented reality, and so on.
61 stars 4 forks source link

framerate question #2

Closed ccqedq closed 2 years ago

ccqedq commented 2 years ago

thanks for your open source, but I have a question: you said all the process run in real-time (approx 40 hz on livox Mid-40) in you paper, does that including Point cloud data collection, and what is the parameter setting of publish_freq when you using livox mid 40 to get a high resolution mapping image for AR marker identification?

yorklyb commented 2 years ago

Hi, we wrote that sentence because we saw such a sentence in LiDARTag's paper: The proposed method runs in real-time and can process data at 100 Hz, which is faster than the currently available LiDAR sensor frequencies. We intended to express the same idea. So the answer is the frequency doesn't contain the time of sampling the point cloud. It's just the time consumption of our proposed method.

Suppose that you use tele-15, then the frequency of our method will be similar but the frequency of the entire process will be faster as the tele-15 lidar can sample faster. We didn't set a publish_freq. As can be seen in our code, there is a stack, and as long as there are enough frames in the stack, we conduct the marker detection and publish the detected result if possible. Suppose that you are using VLP-16, then the size of the stack could be just one. But for livox mid 40, we set the integration_size as 60 (i.e. the size of the stack is 60). Again, suppose that you use tele-15 or horizon, the integration_size can be smaller. The other parameters can be found in the launch file if you are interested.