Deteksi Pelat Nomor Dengan Menggunakan Optical Character Recognition Berbasis Algoritma LSTM

Abstract

Article Info ABSTRACT Article history: Illegal parking is a recurring issue in the Jakarta area, prompting a system transformation that involves adding automatic parking facilities based on building infrastructure and facilities regulations. The development of an automatic parking system can utilize license plate number detection to minimize manual entry of license plate number in the parking system. In this study, LSTM algorithm is trained and implemented on Optical Character Recognition by Tesseract Engine with the aim to create and develop an accurate license plate number detection system to streamline the workflow of institutions requiring vehicle license plate information. The LSTM algorithm demonstrated good performance in detecting license plate numbers with an accuracy rate of 86.36%. However, the LSTM algorithm performance improved significantly to 95.8% accuracy when implemented on Optical Character Recognition by Tesseract Engine. Therefore, based on the evaluation, the LSTM algorithm integrated with Optical Character Recognition is considered the preferred choice for license plate number detection due to its higher accuracy compared to using the LSTM algorithm alone.