Publication: Identification of distance using artificial intelligence
| datacite.subject.fos | oecd::Engineering and technology::Electrical engineering, Electronic engineering, Information engineering::Electrical and electronic engineering | |
| dc.contributor.author | Lim, Yuan Loong | |
| dc.date.accessioned | 2025-05-27T03:19:52Z | |
| dc.date.available | 2025-05-27T03:19:52Z | |
| dc.date.issued | 2024-07 | |
| dc.description.abstract | This report presents the development and evaluation of identification of distance using Artificial Intelligence which utilizing various versions of the You Only Look Once (YOLO) algorithm, with and without the Simple Online and Realtime Tracking (SORT) algorithm, as well as an object distance detection system. The social distancing monitoring system is designed to analyze video feeds and detect violations of social distancing protocols among pedestrians. The effectiveness of five different algorithmic combinations (YOLOv8-L+SORT, YOLOv8-Nano+SORT, YOLOv8, YOLOv3+SORT, and YOLOv3) is assessed based on performance metrics such as Precision, Recall, F1 Score, and efficiency. The results indicate that YOLOv8-only exhibits the highest F1 Score, while YOLOv8-Nano+SORT demonstrates the lowest error rate, suggesting that the inclusion of the SORT algorithm significantly enhances detection accuracy by mitigating false positives. The system's accuracy is evaluated with the performance metrics Mean Absolute Error (MAE) and Root Mean Square Deviation (RMSD) being calculated. The results reveal that the distance estimation performs well with regular-shaped objects but struggles with irregular-shaped objects, highlighting a limitation of using a single webcam for distance estimation. The findings of this research underscore the importance of selecting appropriate algorithmic combinations for specific applications in social distancing monitoring and object distance detection. The integration of the SORT algorithm with YOLO models enhances the accuracy of social distancing violation detection, while the current limitations of the object distance detection system suggest potential areas for future improvement, such as employing multiple cameras or advanced depth-sensing techniques. | |
| dc.identifier.uri | https://erepo.usm.my/handle/123456789/21904 | |
| dc.language.iso | en | |
| dc.title | Identification of distance using artificial intelligence | |
| dc.type | Resource Types::text::report::technical report | |
| dspace.entity.type | Publication | |
| oairecerif.author.affiliation | Universiti Sains Malaysia |