Peiliang
Peiliang
Basically, 2 ways. 1. Collect camera and imu data, then calibrate them offline using kalibr https://github.com/ethz-asl/kalibr 2. refer the datasheet and PCB，then measure them by hand.
1. The camera and imu on iPhone are not axis aligned, so they need extrinsic transform(R, T). 2. No. 3. Yes. So it's challenging to run a vio on general...
The intrinsic parameters of the camera have nothing to do with the camera rotation.
Thanks for the concern. Basically, the sensor driver parts are only included in ViewController.mm, I add some comments for better understanding it, please check the latest commit. You can port...
Hi, yes. Your concern makes sense. Actually, we found in extremely long term datasets(such as the whole campus level), the loop detector will consume too much time. However, for the...
Hi, @loagingray The first thing you need to do is collecting the resized images and calibrating the camera intrinsic parameters rather than using the default one.
Thanks for the report. It's because detecting loop and relocalizing frequently. To avoid this, you can increase the MIN_LOOP_NUM to reduce the loop recall rate. You can also increase the...
Hi, @vonzy VO is not magic, it's hard to get a good result when you give it pure rotation especially when you don't have the accurate extrinsic calibration. Also, the...
sqrt_info is the sqrt of the information matrix, information matrix is the inverse of the covariance matrix. We set 1.5 as the visual measurement noise in the pixel plane.