A method and software for license plate recognition
Вантажиться...
Дата
2024
Науковий керівник
Назва журналу
Номер ISSN
Назва тому
Видавець
National Technical University of Ukraine "Igor Sikorsky Kyiv Polytechnic Institute"
Анотація
The article presents a method for license plate recognition using segmentation through the YOLO detection system combined with a task-oriented approach to training and the use of real-world variable data arrays. The development of metropolises and the constant increase in the number of vehicles on the roads have led to a new level of requirements for road safety systems. Automation, without exaggeration, is the most prioritized direction for the development of these systems. Only through automation can road safety systems process the vast amount of information generated on roads daily. Moreover, automation gradually reduces human involvement in tasks that computational systems can perform with equivalent or greater accuracy. These achievements aim to minimize the influence of the human factor and reduce operational costs. This is particularly important for megacities but also applies to the transportation system as a whole. The purpose of the research is to develop a method for automated license plate recognition to improve the accuracy of road safety systems by reducing error rates, minimizing the excessive use of computational resources during detection, and lowering the cost of such systems. The object of the study is the process of developing automated software systems for ensuring road safety with integrated vehicle identification functionality. To achieve the stated goal, the following objectives were defined: to develop a method for license plate recognition using a task-oriented approach to training combined with the YOLO detection system; to evaluate the impact of prior segmentation of license plates using a specially trained YOLO system on error rates and processing time, as well as to conduct experiments with the proposed training method on real-world images with variable environments to confirm its adequacy. A comparative analysis of the task-oriented training method for the YOLO v5 detection system with the commonly used Optical Character Recognition (OCR)-only approach confirmed the advantages of the task-oriented method for solving license plate recognition tasks. Additionally, the impact of blurring on detection results using the OCR method was investigated. The results of practical research confirm the correctness of the chosen methods for improving the efficiency of license plate recognition.
Опис
Ключові слова
image recognition, image annotation, machine learning, YOLO, license plate recognition, розпізнавання зображень, анотація зображень, машинне навчання, розпізнавання номерних знаків
Бібліографічний опис
Yakovlev, A. A method and software for license plate recognition / Anton Yakovlev, Oleh Lisovychenko // Information, Computing and Intelligent systems. – 2024. – No. 5. – Pp. 125-136. – Bibliogr.: 9 ref.