| Autor: |
Alqoyyum, Muhammad Anas, Wibowo, Adi, Sarwoko, Eko Adi |
| Předmět: |
|
| Zdroj: |
AIP Conference Proceedings; 2023, Vol. 2683 Issue 1, p1-7, 7p |
| Abstrakt: |
Vehicle usage in our lives is increasing exponentially due to the rapid development of the economy. The development of automatic vehicle detection methods through license plates has proven to be one of the solutions in traffic surveillance. This paper proposed applying a two-stage detection method, namely using YOLOv4 (You Only Look Once) and OCR (Optical Character Recognition). With the help of the Tesseract OCR library, recognition results can be done much faster and more effectively when compared to other methods such as backpropagation and SSD. Collection of images of motor vehicle license plates from the Open Images Dataset as the dataset. The YOLOv4 model will be built from 2000 images. The testing result shows that the Mean Average Precision (mAP) scores obtained from the trained model are 89.29%, and the highest Intersection over Union (IoU) score is 73.98% with a precision score of 90%, recall score of 87%, and F1-score 89%. Meanwhile, recognizing license plate serial number characters using Tesseract OCR has good results in bright light conditions. Based on the results, this method has proven to be effective in terms of cost and potency. [ABSTRACT FROM AUTHOR] |
| Databáze: |
Complementary Index |
| Externí odkaz: |
|
