A neuro-based license plate recognition method

number: 
2377
إنجليزية
Degree: 
Author: 
Zaid Mahmoud Mohammed
Supervisor: 
Dr. Ali Abdulhafedh Ibrahim
year: 
2009
Abstract:

Abstract: Automatic License Plate Recognition System (ALPR) developed to automatically recognize and identifies suspected vehicles without the need for a human effort. ALPR system is one of the advanced and recently developed information processing systems to recognize/identify vehicle’s license plate. Practically, identifying a vehicle’s number plate is quite useful in many applications such as: vehicle tracking, traffic control, access control, and many more. Referring to that today’s processors are powerful enough to cope with approaches, which have been impossible for implementation quite recently. In this work, a software program was developed to perform License Plate recognition on Iraqi license plate, this involves the localization of the license plate in the vehicle image snapshot (this is done by digital image processing), plate segmentation (done by morphological image processing), then excluding noisy segments and extracting numeric segments (neural network pattern classifier) and finally recognition of numeric segments (neural optical character recognizer). All car snapshots used in testing the system were restricted to many conditions such as daylight time, fine weather conditions, single lane for passing vehicles, low speed (>= 20km/h) for passing cars. In the developed program, 50 test cars snapshot images were introduced to the program. Results were satisfactory especially from accuracy which was 90% and time cost (average 1.5 seconds) to recognize license plate. Many scopes are available for developing the system such as involving all cities –not only Baghdad- in recognition of plate, making the system to work with video input and many other developing scopes.