About AR Planar
This is based on my previous work on my thesis diploma:
- find all the dominant planes in a point cloud
- merge similar planes into a global map
- measure distances from one plane to the other
- approximate the relative rotation and translation from one pose to the other
Right now the all the point cloud transformations are taken from Tango.
The segmentation and plane merging algorithm is able to accomplish 120 fps on a eight core PC. On Lenovo Phab 2 device it does around 50 fps.
Features:
- plane segmentation
- plane merging in auto mode and manual mode
- project the plane points to the plane
- find distances by averaging the distance from the center of one plane to another
- export the map together with the ADF in scenes with projected points or not
- load previously exported scenes together with relocalization
- record point clounds together with camera poses into binary files and replay them on PC
- as an experimental feature, bullet physics are also supported to throw boxes into walls
The app is also running on PC where the reconstruction algorithm is tested.
Enjoy.