Nowadays, the use of drones has a broader use case. Drones can be used for
both military and non-military purposes, such as data exploration, scientific, commercial
services, research, search and rescue, patrol, and entertainment. Due to the
increasing of the needs of drones and it went with the cheaper price, so that the
drones is produced and assembled with the bad build quality. Therefore the drone is
become more vulnerable to crashes. One of the most common crashes for beginner
users is the crash on landing.
Based on those issues, this thesis proposes vision based precision landing
method for the quadcopter drone equipped with a low-cost camera that is able to
tract 39 x 39 cm of AprilTag and implement a camera-based precision landing
system. The system utilizes the Pixhawk as flight controller, Mission Planner as
the Ground Control Station (GCS) software, OpenMV H7 as the camera sensor,
telemetry as a two-way data stream, and a computer as the control unit, which is
more precise than GPS-based landings and more cost-effective than other alternative
solutions. To ensure the camera-based precision landing system is robust and
reliable, the vision-based precision landing method will be tested in three different
environments, including outdoor, semi-outdoor, and indoor, where each test will be
repeated ten times at an altitude of 2 and 3 meters.
The results obtained that after testing, the system indicate that the drone with a
camera-based precision landing system is capable of autonomously scanning, tracking,
and landing on a landing target. The average accuracy for 2 meter altitude in
the outdoors is 8.94 cm, semi-outdoor 9.1 cm, and indoor 11.7 cm. Meanwhile, the
average accuracy for 3 meter altitude in the outdoors is 9.35 cm, semi-outdoor 9.45
cm, and indoor 10.8 cm.