Recognition Of Malagasy Vehicle License Plates Using Hough Transform And Multilayer Perceptron For Automated Garage Door Opening

Abstract

Automatic License Plate Recognition (ALPR) has become a critical component in modern intelligent transportation systems, yet its deployment in resource-constrained environments presents unique challenges. This study investigates the design and implementation of an automatic license plate recognition system specifically intended for garage door opening automation within the Malagasy context. By integrating classical image processing techniques—namely Canny edge detection and Hough transform for plate localization—with a multilayer perceptron (MLP) neural network for character classification, the proposed system achieves a balance between computational efficiency and recognition accuracy. An empirical evaluation was conducted using a dataset of static vehicle images captured under varying lighting conditions. The results demonstrate a plate localization success rate of 71.42% and a character recognition accuracy of 99% for previously learned plates, yielding an overall system efficiency of 85.21%. Furthermore, the study compares the performance of noise-injected training versus standard training, revealing that noisy training significantly enhances robustness, reducing error rates under progressive input degradation. Despite limitations in handling inclined planes and complex backgrounds, the proposed framework provides a practical, low-cost solution for automated access control. The study concludes by outlining future directions, including the integration of Fourier descriptors for rotation invariance and the development of real-time capabilities for nighttime and adverse weather conditions.