This project can be split into 3 phases.
Phase 1 aims at developing realtime and robust visual feature detector and tracker for beagleboard.
Have been working on porting OpenSURF to DM642 for a while and could achieve real-time performance, so this won’t be too difficult.
Phase 2 will focus on SLAM based on feature detection/tracking.
Phase 3 will integrate this module into open source quadrotors like MikroKopter.Pose estimation will use both visual and gyro/accelerometer feedback.Most algorithms will run in BeagleBoard instead of original MCUs these projects are using.