License Plate Detection, Recognition
and Automated Storage
students' project

Project Summary

The project is aimed at designing a prototype system for automated detection of country identification sticker on an observed vehicle, license plate location and reading. The collected information is to be stored in a database, together with the accompanying information (e.g. time, date and place of passing the border).

The prototype system will use a knowledge base containing the information about the license plate syntax used in different countries. The system will first try to locate the country identification sticker in the image of the observed object and to identify the country of origin. Subsequently the license plate will be searched for, and the identification characters will be recognized and stored in the database. The system should be robust and able to identify the country of origin even with no visible identification sticker, based on the syntax of the license plate characters. The system should be robust also with respect to illumination conditions and noise (e.g. dirty license plate, bad visibility). The project will be carried out with a group of students (4. year) attending the classes "Pattern recognition" and "Intelligent systems", and who are working towards their diploma theses under the supervision of prof. S. Ribarić. The project brings together knowledge from several areas: digital image processing, pattern recognition, artificial intelligence, knowledge representation and reasoning.

The project will be carried on in several phases:

• Phase I
  • building test image base containing images of different vehicles, taken from the rear and under different light conditions,
  • design of algorithms for gray-scale image enhancement (preprocessing stage),
  • web page design;
• Phase II
  • development of algorithms for image thresholding with adaptive global/local thresholds,
  • design of algorithms for gray-scale image enhancement (continued),
  • image segmentation aimed at locating the country identification sticker and the license plate;
• Phase III
  • application of active vision for fast detection of the country identification sticker and the license plate,
  • building the knowledge base with the syntax rules for license plate codes for different countries;
• Phase IV
  • design of OCR (optical character recognition) subsystem based on multi-layered perceptron and selection of robust methods for automated reading of alphanumerical codes on license plates,
  • exhaustive testing of prototype system and evaluating the recognition results,
  • design of demo programs to be available on the web site.

The prospective users of this system are all institutions wishing to automate the vehicle surveillance and collecting the license plate data (border passes, parking lots, transportation companies, bus terminals, government, military etc.)

In our opinion, the project has multiple benefits as it motivates the students for real-world project assignments (at the end of their study) which include advanced technologies. On the other side, it embodies the efforts to create sophisticated products ... based on national knowledge. Furthermore, the participation in the project will enable the future young experts the collecting of new knowledge and skills, bringing together knowledge from areas of digital image processing, pattern recognition and artificial intelligence. The prototype system has also its practical value as it can be (under additional funding) installed at border passes and used for counting vehicles (cars, trucks) coming from different countries. Large transportation corporations or bus terminals could also find the system useful. The system could be upgraded by statistical data analysis tools and used for various applications (giving reports about traffic intensity distribution, border crossing bottleneck prediction, regional evaluation of tourist season success etc.)

The expected final product is a prototype of the system for automated detection, recognition, reading and storing of country identification and license plate codes. It will include:

• image database containing images of cars and trucks;
• module for image preprocessing and detection of country identification sticker (plate);
• module for detection of license plate location in the image;
• knowledge base with syntax rules of license plate codes used in different countries;
• module for automated reading and storing of license plate codes;
• module for increasing the robustness - it should be activated when the system encounters difficulties due to license plate dirtiness or bad illumination conditions;
• database generated by using the test image base;
• executable code (demo) which could be tested by users with their images (taken by a digital camera);
• recognition results (%).

Copyright(c) Department of Electronics, Microelectronics, Computer and Intelligent Systems
Faculty of Electrical Engineering and Computing
University of Zagreb

Zoran Kalafatić
Last change: 10. 11. 2003.