Automatic system for localization and recognition of vehicle plate numbers

 

N. Vázquez, M. Nakano & H. Pérez-Meana

 

Graduate and Research Section Mechanical and Electrical School, Culhuacán Campus National Polytechnic Institute of Mexico. Av. Santa Ana No. 1000, San Francisco Culhuacán, 4430 México D. F. México. E-mails: ntecpanecatl@hotmail.com; hmpm@calmecac.esimecu.ipn.mx; hmpm@prodigy.net.mx

 

Received: April 18^th 2001.
Accepted March 6^th 2002.

 

ABSTRACT

This paper proposes a vehicle numbers plate identification system, which extracts the characters features of a plate from a captured image by a digital camera. Then identify the symbols of the number plate using a multilayer neural network. The proposed recognition system consists of two processes: The training process and the recognition process. During the training process, a database is created using 310 vehicular plate images. Then using this database a multilayer neural network is trained to identify the symbols in the vehicles plate. While the recognition process consists of four stages: The number plate localization stage, the binarization stage, the segmentation stage and the recognition stage which uses the previously trained multilayer neural network The performance of proposed system is evaluated using more than 1200 symbols from the 310 captured images. The simulation results show that approximately 91.5% of the 310 plate images in the vehicle have been correctly located. The proposed system performance, regarding the identification of numbers and letters in the plate, was evaluated separately. Here the recognition rate is 95.55°% and 91.6%, respectively. So the global recognition rate of the vehicle number plate becomes approximately 91.2%. Then from the simulation results it follows that the proposed system works fairly well and then it may be applied in the solution of several practical problems that require automatic number plate identification.

Keywords: Alphabet recognition, Number plate, Identification with neural networks.

 

RESUMEN

Se propone un sistema de identificación de placas vehiculares que facilite y agilice la identificación de las mismas a través de redes neuronales, una vez que han sido obtenidas las características de la placa por medio de una imagen tomada con una cámara fotográfica digital. El sistema propuesto consiste de dos procesos: El proceso de entrenamiento y el proceso de reconocimiento. El proceso de reconocimiento consiste en la localización de la placa dentro de la imagen capturada, la binarización de la misma, la segmentación de los símbolos por medio de la técnica de etiquetamiento, la codificación de los símbolos segmentados y el reconocimiento de los mismos usando las redes neuronales previamente entrenadas por el proceso de entrenamiento. El proceso de entrenamiento por su parte consiste de la creación de la base de datos y el entrenamiento de las redes neuronales multicapas. El funcionamiento del sistema global se evaluó usando el porcentaje de acierto de reconocimiento de los símbolos (números y letras) de las placas correspondientes a 310 imágenes capturadas. Los resultados obtenidos muestran que aproximadamente en un 91.5% de las imágenes se han localizado correctamente la posición de la placa. Por su parte el porcentaje de acierto en el reconocimiento de los dígitos y letras en la placa, se estimaron separadamente, obteniéndose porcentaje de reconocimiento de aproximadamente 95.5% y 91.6% respectivamente, mientras que el reconocimiento global de las placas consistentes de 3 números y 3 letras es de 91.2%. De los resultados obtenidos podemos concluir que el sistema propuesto funciona acertadamente y podría ser empleado en diversos sistemas que requieran detección automática de placas.

 

DESCARGAR ARTÍCULO EN FORMATO PDF

 

References

[1] Web page: www.pgjdf.gob.mx

[2] Chutape, L. Li, and Xiaodong Q., Automatic License Number Extraction And Its Parallel Implementation, Proceeding of The 25^th Annual Conference of the IEEE, Industrial Electronic Society, IECON'99, Volume 2, pag. 704-709, 1999.         [ Links ]

[3] Zinino R. and Rovetta S., 1999 Visual Location Of License Plates by Vector Quantization, Proceedings of the 1999. IEEE International Symposium on Circuits and Systems, ISCAS'99. Vol. 4, pag. 135-138.         [ Links ]

[4] Zimic N., Ficzko J., Mraz M., Virant J., 1997 The Fuzzy Logic Approach to the Car Number Pate Locating Problem, Proceedings of Intelligent Information System, IIS'97, Pag. 227-230.         [ Links ]

[5] Coetzee C., Botha C. and Weber D. 1998. PC Based Number Plate Recognition System, Proceedings of IEEE International Symposium on Industrial Electronics, ISIE98, Vol.2, Pag. 605-610.         [ Links ]

[6] Raus M., Kreft L., 1995 Reading Car License Plates by the Use of Artificial Neural Network, Proceedings of The 38^th Midwest Symposium on Circuits and Systems, Proceedings, Vol.1, Pag. 538-541.         [ Links ]

[7] Maged M. Fahmy M., 1994 Automatic Number-Plate Recognition: Neural Network Approach, Proceedings of Vehicle Navigation and Information Systems Conference, Pag. 99-101.         [ Links ]

[8] de Vena L., 1993 Number Plate Recognition by Hierarchical Neural Network , Proceedings of 1993 International Joint Conference on Neural Network, Vol.3, Pag. 2106-2108.         [ Links ]

[9] Gendy S., Smith C. L., and Lachowicz S., 1997 Automatic Car Registration Plate Recognition Using Fast Hough Transform, Proceedings of International Carnahan Conference on Security Technology, Pag. 209-218.         [ Links ]

[10] He M. G., Harvey A. L., Vinay T., 1996 Hough Transform in Car Number Plate Skew Detection, Proceedings of Fourth International Sympsium on Signal Processing and Its Application, ISSPA 96, Vol. 2, Pag. 593-596.         [ Links ]

[11] Garris M.D., Wilson C.L., Blue J. L., 1998 Neural Network-Based Systems for Handprint OCR applications, IEEE Transaction on Image Processing Vol. 8, No. 8, pag. 1097-1112.         [ Links ]

[12] Mori S., Suen C.Y., Yamamoto K., 1992 Historical Review of OCR Research and Development, Proceedings of the IEEE Vo. 80, No. 7, pag. 1029-1058.         [ Links ]