Image processing framework for traffic monitoring

This paper describes an image processing framework capable of tracking moving objects in two or three dimensions. The objects are observed by multiple cameras and their path can be recorded without markers, special cameras or background. This paper shows how position coordinates can be acquired from the image of objects. The moving regions of a frame are specified by the difference between the frame and an adaptive background. The adaptive background is calculated from a sequence of frames, meaning that the program can also work in changing light conditions and can handle slow-moving objects. Afterwards, the identified moving regions are tracked inter-frame. The positions of objects received from cameras are then converted to a common word space coordinate system. At this point, the position coordinates are ready for further processing. The program uses a wide array of technological tools, such as hardware acceleration, multi-threaded processing, three dimensional display of objects and network data transmission among the cameras. An integrated framework has also been developed, which is easily customizable, and can be adapted to specific tasks.