liquidcronos / optical-stabilisation

ROS Package to estimate a UAVs velocity using a monocular Camera and a ultrasonic sensor
0 stars 0 forks source link

startdistance jumps #1

Closed liquidcronos closed 6 years ago

liquidcronos commented 6 years ago

the classification of immobile objects fails sometimes due to a few markers which stay active despite the large jumps the make (those are likely caused by low lk levels and a small window size, this however should present no problem as long as they are properly filtered)

liquidcronos commented 6 years ago

long trails seem to happen when multiple features move close together. This can be partially fixed by increasing the minimum distance between features

liquidcronos commented 6 years ago

adding status==True as requirement for good points fixes the jumping point problem, the long trails remain however

liquidcronos commented 6 years ago

constraining maximum flow instead of distance solves the problem (function tomuchvel).

liquidcronos commented 6 years ago

problem was only solved due to error in tomuchvel and still persists.

liquidcronos commented 6 years ago

Fixed, error was due to missing np.abs in tomuchvel