Publication: Implementation of feature-based car plate detection
| datacite.subject.fos | oecd::Engineering and technology::Electrical engineering, Electronic engineering, Information engineering::Electrical and electronic engineering | |
| dc.contributor.author | Tai, Yung Ern | |
| dc.date.accessioned | 2025-05-19T08:50:29Z | |
| dc.date.available | 2025-05-19T08:50:29Z | |
| dc.date.issued | 2024-08 | |
| dc.description.abstract | Automatic Number Plate Recognition (ANPR) systems have witnessed significant advancements in recent years, fuelled by the convergence of algorithmic innovations and technological enhancements. ANPR consists of license plate detection, character segmentation and character recognition. While artificial intelligence and convoluted neural networks are often used for license plate detection systems, they require training and may be computationally expensive. Thus, the objective of this project is to develop a license plate detection algorithm using computer vision and image processing techniques, leveraging edge detection as a feature based approach. The second objective is to evaluate and analyse the performance of the developed algorithm in terms of its accuracy and processing times using a factorial experiment design. Lastly, the algorithm is compared with previous work. Thus, two license plate detection algorithms were developed using C++ and OpenCV, one using canny edge detection and another employing Laplacian of gaussian (LOG) edge detection. Image processing techniques such as applying grayscale, filtering and contour analysis were also used. LOG edge detection achieves a higher accuracy than canny thus it would be used for the factorial experiment. 9 experiments were conducted with two independent variables identified to be the LOG Sigma and median filter size, each with three values. After that, two optimal set of values were identified: LOG sigma 1.4 and no median filter produced an accuracy of 48.61% with processing time of 1043.887ms and LOG Sigma 1.6 with median filter size of 5×5 pixels produced an accuracy of 51.39% and a processing time of 12346.7ms. A chi-square test of independence on accuracy suggests that the LOG Sigma affects the accuracy while the median filter does not. A one-way ANOVA for the processing time suggests that the median filter size significantly affects the processing time while the LOG Sigma does not. The algorithm was compared with previous work and highlights the importance of processing time as a metric for hardware implementation. | |
| dc.identifier.uri | https://erepo.usm.my/handle/123456789/21728 | |
| dc.language.iso | en | |
| dc.title | Implementation of feature-based car plate detection | |
| dc.type | Resource Types::text::report::technical report | |
| dspace.entity.type | Publication | |
| oairecerif.author.affiliation | Universiti Sains Malaysia |