Note: This content is accessible to all versions of every browser. However, this browser does not seem to support current Web standards, preventing the display of our site's design details.


Multi-Perception Vision Control


R. Hofmann

Master Thesis, HS13 (10287)

Pan-tilt-zoom (PTZ) cameras are a key element in the area of camera surveillance. Commercially available PTZ units usually have built in modes like patrolling between prede ned setpoints or tracking of moving objects. Zoomed in PTZ cameras only deliver information of a narrow eld of view while the rest of the area is not monitored. Static wide angle cameras on the other hand always monitor the same area but the image quality is typically not sucient for tasks like identication of individuals. Given the benets and handicaps of both PTZ and static wide angel cameras this thesis' aim was to develop a system which combines a dynamic PTZ camera with a static wide angle camera such that they work together. In this setup the wide angle camera provides overview information to which the PTZ camera is moved. Besides several tasks in surveillance possible applications can be found in computer controlled sports broadcasting or other entertainment tasks. Using a pan-tilt-unit designed in a former project and two high quality industrial Ethernet cameras a prototype of such a multi perception vision system was built. Being it a portable system which is used in dierent scenarios a semiautomatic computer vision based calibration process was implemented. The calibration process outputs a continuous mapping function which maps the image coordinates from the wide angle camera to pantilt angle coordinates of the PTZ camera. Given the mapped coordinates of a desired position an optimal trajectory for moving the PTZ camera is generated using tools from model predictive control. A graphical user interface was created for handling the user interaction and displaying the images of the cameras. To demonstrate a variety of possible applications a manual operating mode was implemented along two automatic modes based on computer vision. The calibration can be achieved to an accuracy below 0:1 solely using the camera images in 5 to 10 minutes. In manual mode the PTZ camera is moved to the position of a mouse click of the user in the wide angle camera image. In automatic mode the PTZ camera obtains high resolution images of faces detected in the wide angle camera image or it tracks an object based on its color. Multi perception vision systems can be used for a grate variety of computer controlled surveillance or entertainment tasks. The implemented way of connecting the information of the wide angle camera to movements of the PTZ camera using a continuous mapping function proved to be reasonable and accurate. The performance of the presented prototype needs to be improved by redesigning the hardware.


Type of Publication:

(12)Diploma/Master Thesis

N. Kariotoglou

File Download:

Request a copy of this publication.
(Uses JavaScript)
% Autogenerated BibTeX entry
@PhdThesis { Xxx:2014:IFA_4783
Permanent link