Automated License Plate Recognition systems (ALPR), take an important role in modern traffic control systems , as well as for many security purposes.
Abstract
Automated License Plate Recognition systems (ALPR), take an important role in modern traffic control systems , as well as for many security purposes. Typically, such systems are static, and operate under restricted conditions, such as fixed illumination and stationary background. Furthermore, most systems use special cameras, sometimes a set of several cameras with built-in illumination and a related computer. Also, such system must usually be installed and calibrated, and tend to be very expensive.
Our algorithm receives a video input from a smartphone camera, finds and cut license plate out of the frames, separates the numbers from each other, and finally classifies the digits and saves them as output in a text file.
The algorithm integrates methods from different fields such as pattern recognition, geometric filtering, analysis of edge statistics and several techniques from learning theory, in order to achieve good performance while maintaining short enough running time for working in real time.
The algorithm was programmed using Matlab software, using few of its many toolboxes such as Optimization, Statistics, Signal Processing and Image Processing toolbox. The results we got are much better than the ones we expected, and therefore our next objective will be implementing the algorithm on a smartphone, in order to check it’s capabilities in a real time system.