| Abstract | Nowadays, object recognition methods are proposed from many researcher. Multi-class
object recognition system is also interested and developed. In robotic, robot has to
autonomous make decision from what it has just seen. Like humans, we firstly learn
about many objects that contain in eye view. Then, we can use this information to
make a decision in continue doing job. In the same case, robots need the system for
object recognition from its own view in order to process the next task to do. Multi-class
object recognition is needed to help robot in making decision. In my thesis, I try to
explore the appropriate algorithm to work for object classification from aerial imagery.
I focus on a few instances of object in order to classify such as car, building, road and
tree with grass. I try to choose appropriate method to make system run as near as real
time processing. Also, this method should give the accuracy of result to be satisfied
enough. Then, I test this algorithm with data which are collected from real mission
in UAVs system. The challenge is that we should use information from the top view
scene which might have less information thatn side view. Also, image which are got
from UAVs system are not quite good in resolution and clear enough. |