PENGENALAN OTOMATIS PELAT NOMOR KENDARAAN BERMOTOR DALAM CITRA DIGITAL MENGGUNAKAN JARINGAN SARAF TIRUAN BACKPROPAGATION UNTUK IDENTIFIKASI DAN KLASIFIKASI YANG AKURAT DALAM BERBAGAI APLIKASI, SEPERTI MANAJEMEN PARKIR, PEMANTAUAN LALU LINTAS, DAN SISTEM KEAMANAN KENDARAAN

Abstract

Automatic recognition of motor vehicle license plates in digital images is crucial for applications such as parking management, traffic monitoring, and vehicle security. This study implements a backpropagation artificial neural network to enhance accuracy in identifying and classifying license plate characters. The research methodology includes image acquisition, preprocessing using grayscale conversion, noise reduction, and edge detection to improve image clarity. Feature extraction isolates key characteristics, which serve as input for the neural network. The backpropagation algorithm adjusts weights and biases through iterative learning to minimize errors. Testing on 100 image samples demonstrated an overall accuracy of 88%, with well-lit images achieving 94% accuracy, while noisy images had a reduced accuracy of 76%. Errors primarily resulted from poor lighting, distorted characters, and occlusions. The findings indicate that optimized dataset training, advanced preprocessing techniques, and refined neural network parameters significantly improve recognition performance. The system offers an efficient alternative to manual identification, reducing human error and improving operational efficiency. Future work includes integrating deep learning models, increasing training data diversity, and optimizing network layers to enhance robustness across various environmental conditions. This research contributes to automated intelligent transportation systems, offering real-time, high-accuracy vehicle identification for enhanced security and traffic regulation.