Nowadays, detection and analysis of moving objects and patterns are getting more and more important role in the processing of visual images and video sequences. The processing of visual information is represented now in our everyday life and there is an extensive research activity in several fields like in medical sciences, astronomy, industrial image processing, security engineering and in traffic safety.
Object detection and pattern recognition on images recorded with digital cameras mounted on moving vehicles is a challenging task, which requires selection of efficient and robust detection and tracking algorithms besides the powerful and fast electronic tools and computers implemented for the actual conditions.
With help of various motion-estimation algorithms based on the image extracted information data several image features can be recognized, the interesting objects can be tracked and under given conditions even their future positions are predictable in advance.
Object detection itself a moving platform is a challenging task, even if the kinetic data are known from other sources. Estimates based on the changing pictures makes the task interesting. The goal is to find such patterns and motion features, where sufficient information can be obtained for the applicable evaluation.
In my thesis I briefly summarize the image processing methods applicable for pattern detection, recognition and moving object tracking. For reconstruction of 3D environment on the basis of 2D image frames calibration of digital camera has been done. Using the intrinsic calibration parameters of this camera I carried out a continuous estimation for the varying distance between the moving camera and the tracked moving car.
Another main task of my thesis was detection and tracking of moving pedestrians on video-sequences shot from a moving vehicle platform with help of image foreground-background separation processing methods.