[Studies in Computational Intelligence 481] Artur Babiarz, Robert Bieda, Karol Jędrasiak, Aleksander Nawrat (auth.), Aleksander Nawrat, Zygmunt Kuś (eds.) - Vision Based Systemsfor UAV Applications (2013, Sprin

58 Z. Kuś and A. Nawrat
camera operator. Utilizing such an approach requires the operator to observe camera
image on the screen which is located on the ground. The video is streamed via
a communication system like [33] or [35].
However, this approach is considerably limited by the manual skills of the operator.
This limitation is crucial in case of the occurrence of flight trajectory disturbances.
An example of such a disturbance is strong gust of wind. It results in
the object loss from the operator's screen and the operator cannot track the object
any longer. Moreover, the same problem occurs when we use methods based on
the image processing. These methods [1, 2, 3, 4, 5, 6, 7] are very effective when
the tracked object occurs in the camera field of view. Rapid disturbances can
make that the tracked object is out of the picture. When the tracked object movements
are not the reason that we can not see the object then the image processing
methods [14] – [23] do not give us any valuable information. The aforementioned
situation is examined in the following paper. Hence the automatic object
tracking method for rapid camera movements will be proposed herein.
2 Camera Movements Correction in 3D Space
The following section will focus on the basic elements of the solution of the problem
presented in the introduction. The 3D modeling of a helicopter equipped with
the camera is far more complex; however, it considerably increases the possibilities
of camera control.
Below there are presented basic assumptions concerning geometrical relations
for the examined problem:
a) the helicopter position equals camera position (displacement of the optical center
of the camera lens in relation to the center of mass of the helicopter is minor
and easy to take into account);
b) the camera orientation is a sum of helicopter orientation and camera head rotation
(we do not model helicopter behavior, but we utilize a dynamic model of
the impact of helicopter rotations on a camera - taking into consideration that
the helicopter has its own flight trajectory control system e.g [26], [27]);
c) the tracked object moves on a flat ground 0;
d) prior to commencing automatic object tracking the helicopter is manually directed
at the area where the tracked object is situated. Next, the camera working
in manual mode is directed at the object tracked by the operator. Then the operator
marks the object on a screen or the object is located by image processing
algorithms;
e) automatic object tracking is commenced when the object is in the center of the
camera field of view;
f) the camera head may rotate within the range of 360 degrees in horizontal plane -
parallel to the bottom of helicopter.
g) the camera is always above the object;
h) the helicopter with the camera mounted at the bottom of a helicopter body flies
in a position approximate to horizontal;