Real-time detection, tracking and classification of multiple moving objects in UAV videos
| Autor: | Huseyin Can Baykara, Erdem Biyik, Gamze Gul, Deniz Onural, Ahmet Safa Ozturk, Ilkay Yildiz |
|---|---|
| Jazyk: | angličtina |
| Rok vydání: | 2017 |
| Předmět: |
Computer science
0211 other engineering and technologies Image registration ComputerApplications_COMPUTERSINOTHERSYSTEMS 02 engineering and technology Object tracking Tracking (particle physics) Automation 0202 electrical engineering electronic engineering information engineering Computer vision Low altitude 021110 strategic defence & security studies Aerial image classification Surveillance Scope (project management) business.industry Track (disk drive) Aerial video Deep learning Moving object detection Object detection Transfer learning Real time 020201 artificial intelligence & image processing Artificial intelligence business |
| Zdroj: | Proceedings-International Conference on Tools with Artificial Intelligence, ICTAI 2017 ICTAI |
| Popis: | Date of Conference: 6-8 Nov. 2017 Conference name: IEEE 29th International Conference on Tools with Artificial Intelligence (ICTAI), 2017 Unnamed Aerial Vehicles (UAVs) are becoming increasingly popular and widely used for surveillance and reconnaissance. There are some recent studies regarding moving object detection, tracking, and classification from UAV videos. A unifying study, which also extends the application scope of such previous works and provides real-Time results, is absent from the literature. This paper aims to fill this gap by presenting a framework that can robustly detect, track and classify multiple moving objects in real-Time, using commercially available UAV systems and a common laptop computer. The framework can additionally deliver practical information about the detected objects, such as their coordinates and velocities. The performance of the proposed framework, which surpasses human capabilities for moving object detection, is reported and discussed. ACKNOWLEDGMENTS We thank Ersin Yar, Bumin Kaan Aydın, Dr. Cem Tekin and Dr. Orhan Arıkan for comments and fruitful discussions throughout the project. This work was supported by HAVELSAN Inc. and Bilkent University, Electrical and Electronics Engineering Department. We acknowledgefunding by TÜB˙TAK B˙DEB 2209-B grant. |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|