Vehicle detection in aerial images

International Scientific Conference «AGRITECH-2019: Agribusiness, Environmental Engineering and Biotechnologies»

农业科学;生态环境科学;生物科学

Dorrer, Georgy^1^2 ; Koriukin, Maksim^1 ; Yushkova, Svetlana^1 ; Sviridova, Lidiia^1

Reshetnev Siberian State University of Science and Technology, 31 Krasnoyarskiy Rabochiy Pr., Krasnoyarsk, Russia^1

Siberian Federal University, 26 Kirenskogo St., Krasnoyarsk, Russia^2

关键词: Aerial images;    Complex background;    Convolutional neural network;    Environmental conditions;    Ground truth;    Scene complexity;    Vehicle detection;   

Others  :  https://iopscience.iop.org/article/10.1088/1755-1315/315/2/022014/pdf
DOI  :  10.1088/1755-1315/315/2/022014

来源: IOP

【 摘 要 】

The vehicle detection in the aerial images is widely used in many applications. Comparing with the object detection in the ground view images, vehicle detection in the aerial images remains a challenging problem due to the small size of vehicles, monotone appearance and complex background. In this paper, we propose the solution of this issue using the convolutional neural networks. We further introduce the large-scale vehicle detection dataset with ground truth annotations for all the vehicles in the scene that considers the scene complexity due to the environmental conditions. We show the performance of the trained model with other popular neural work architectures.