Can you help me understand whether you've used the minimum number of possible April tags? what would that be for navigation in a room of the size you mentioned?
Hey Bill, thanks for the question. In the setup above, I was using a single camera on the QCar (The RealSense D435). As the car drives around, that camera might lose sight of the tag (the RGB lens has a 69.4 degree horizontal FOV or field of view only). To maximize my chances of always seeing a tag (for example when I turn the car in another direction) I decided to use multiple tags. If my camera had a wider FOV, I might have been able to get away with fewer tags. For example, using the single front facing CSI camera on the QCar (with a 160 degree FOV) might theoretically allow me to get away with 3 tags. For an alternate setup, if I had decided to use multiple cameras (the 4x CSI cameras on the QCar can provide a 360 degree FOV together), I would have been able to use a single tag. No matter how the QCar would turn, at least 1 of the cameras would detect the tag and register the camera pose. To summarize, the algorithm itself is quite elegant and allows you to use a single tag. However, the number of tags you would need in your own setup is a function of numerous factors, such as room size, the field of view of your camera, the number of cameras you are using etc..
Can you help me understand whether you've used the minimum number of possible April tags? what would that be for navigation in a room of the size you mentioned?
Hey Bill, thanks for the question. In the setup above, I was using a single camera on the QCar (The RealSense D435). As the car drives around, that camera might lose sight of the tag (the RGB lens has a 69.4 degree horizontal FOV or field of view only). To maximize my chances of always seeing a tag (for example when I turn the car in another direction) I decided to use multiple tags. If my camera had a wider FOV, I might have been able to get away with fewer tags. For example, using the single front facing CSI camera on the QCar (with a 160 degree FOV) might theoretically allow me to get away with 3 tags.
For an alternate setup, if I had decided to use multiple cameras (the 4x CSI cameras on the QCar can provide a 360 degree FOV together), I would have been able to use a single tag. No matter how the QCar would turn, at least 1 of the cameras would detect the tag and register the camera pose.
To summarize, the algorithm itself is quite elegant and allows you to use a single tag. However, the number of tags you would need in your own setup is a function of numerous factors, such as room size, the field of view of your camera, the number of cameras you are using etc..
How can I use in python ?
Hi thanks for watching. Please leave your message to info@quanser.com, we will reply through emails.