Estimating vehicle speed using image processing

Abstract

This research uses image processing to estimate car speed. Given a sequence of real-time video of traffic images. The images are converted to monochrome images then edges are extracted and then quantified the resulting images to classify the objects and find the cars, after detecting the cars, they can be tracked in different frames and their speed can be estimated by calculating the position of the car object in different frames and compare it to a ground truth. This research uses a different approach to estimate speed, rather than using a reference object to estimate speed, the Matlab function is used and different relationships between the objects that are extracted from the image itself. This method does not need camera calibration and does not need any extra equipments other than a video camera attached to a computer and can be used by traffic police to monitor highways car or normal roads, and can be combined with a system to detect the license plate number to create a complete system to issue an automatic speed ticket for drivers who exceeds the speed limits. The system is tested in different roads and the results show the validity of the used method.