The rise in the speed of the traffic observation systems can be accredited to the exponentially growing performance of computing systems. The improvement of the last years even gives us the opportunity to analyse the videos recorded by cameras. The methodology of image processing has been known since several years, but the storage of proper amount of pictures with proper resolution in terms of the observation required a huge memory capacity and a fast processor, which made the real-time analysing of the pictures nearly impossible. This slowness of the observation systems left us with the remaining possibility of ’human’ observation. Even a short period of traffic observation can produce a huge amount of data. For the processing software it is a significant problem to sort out the neccessary sub-elements for the actual project, even if the parameters are correct and all of the important traits which are required e.g. for the traffic management or police surveillance of a city, are present on the pictures.
From the huge amount of information contained by the pictures, this dissertation studies the recognition of moving vehicles on public roads. The repeatibility and the controllability of the events are enabled with fixed video records. After recording the videos of the oncoming traffic from an overpass above a highway, they were analysed thoroughly. The first step in the recognition process is to subtract the background from the foreground. To subtract the background the Gaussian Mixture probabilistic model was used. The observation of the moving objects could be started only after the identification of the real observation area. First, I got acquainted with several algorithms for detecting moving objects, then I tested some of these algorithms and compared them by precision and speed. For this, both the black and white and color versions of the algorithms were used and rated. I used the OpenCV which is an open source computer vision library of programming. To code the traffic analysing program the C++ language was chosen, because it has the fullest OpenCV library. OpenCV contains more than 500 functions from which some are optimalized to the given processor, and these functions provide the opportunity to create a more complex program package (e.g. detecting irregularities, speed measuring) in the future.