HKUST-Aerial-Robotics / VINS-Mobile

Monocular Visual-Inertial State Estimator on Mobile Phones
GNU General Public License v3.0
1.27k stars 524 forks source link

Influence of Inverse Depth #112

Open xushangnjlh opened 6 years ago

xushangnjlh commented 6 years ago

Thanks for your wonderful work. I noticed that the depth (not inverse depth) of mappoint varies a lot during exploration, leaving the map size of vio system not very consistent or deviating from real world. I am wondering, if change the parametrization of mappoint to usual x,y,z, whether the system can have a comparative performance, as the scale estimation is not precise when initialized? Ref. https://github.com/HKUST-Aerial-Robotics/VINS-Mono/issues/118

PeiliangLi commented 6 years ago

Yes, it will at least have a comparable performance.