To search, Click below search items. |
| All Published Papers Search Service |
|
|
|
Title |
A Systematic Review of Lightweight Anomaly Detection Techniques for UAV Video Surveillance. |
|
Author |
Wejdan Alotaibi and Dr. Ohoud Alharbi |
| Citation |
Vol. 26 No. 2 pp. 59-64 |
|
Abstract |
Recent advancements in aerial surveillance have intensified interest in UAV-based anomaly detection, particularly as real-time monitoring becomes increasingly critical in dynamic environments. This systematic review synthesizes sixteen peer-reviewed studies published in 2025 that investigate deep learning techniques tailored for UAV anomaly detection. The search process, conducted through the Web of Science Core Collection and augmented with snowballing, yielded three dominant research themes: lightweight real-time detection models, spatiotemporal and self-supervised anomaly learning, and robustness under challenging environmental conditions. The findings highlight a clear shift toward computational efficiency through compact CNN architectures and hardware-accelerated designs, alongside improved temporal reasoning via LSTM-based and prediction-driven frameworks. Additionally, several studies emphasize enhancing visibility and structural clarity under low-light, noisy, or variable-scale conditions through multimodal fusion and multi-scale refinement. Despite these advancements, limitations remain concerning dataset diversity, standardized evaluation, and real-world experimentation. The review concludes by outlining future directions that include expanding UAV anomaly datasets, strengthening temporal modeling, and validating models through practical deployments. |
|
Keywords |
UAV anomaly detection; lightweight deep learning; spatiotemporal modeling; self-supervised learning; multimodal fusion; multi-scale refinement; aerial surveillance; systematic review. |
|
URL |
