An Algorithm for License Plate Recognition System for Non-stop Toll Stations

License plate recognition system (LPRS) is the hard core of the intelligent traffic system. In this paper, license plate (LP) frame was captured from video image sequences by background-difference, and an improved Kalman filter method was adopted to update the background. By means of mathematical morphology (MM) and edge characteristic analysis, LP orientation was conducted. After the binarization and adoption of multi-indexes evaluating function for color reverse, characters were segmented via connected components analysis. Methods employed in this paper, such as the improved Kalman filter for updating background, the binarization based on gray scale difference between background and characters in LP, the color-reversing judgement function, etc, are novel and efficient. With samples captured from videos under different illuminating conditions, the experiment indicates that it is feasible to apply the algorithm to LPRS.