Attributable to global economic development and rapid urbanization, traffic explosion has been observed. This situation becomes very prominent as the traditional traffic control systems are incapable to efficiently monitor and control the traffic. Therefore, this paper presents automatic vehicle detection from satellite images using deep learning approaches. For this purpose, two most renowned and widely used detection algorithms (YOLOv4 and YOLOv3) have been employed to develop satellite image vehicle detector using publicly available DOTA dataset. This work confirms the supremacy of YOLOv4 over YOLOv3 by large improvements in mAP, IOU, precision, recall, F1-score with increase of 45%, 20%, 11.1%, 45.9%, and 29.5%, respectively. Therefore, these investigational results verify the strength of YOLOv4 algorithm for satellite image vehicle detection and recommend its use for the development of an intelligent traffic control system.
Automatic Vehicle Detection from Satellite Images Using Deep Learning Algorithm
Advs in Intelligent Syst., Computing
2021-06-27
12 pages
Article/Chapter (Book)
Electronic Resource
English
Vehicle Pollution Detection from Images Using Deep Learning
Springer Verlag | 2020
|Vehicle Detection from Satellite Images
Transportation Research Record | 2009
|Automatic Detection from a Satellite Using Hyperspectral Images
British Library Conference Proceedings | 1998
|