Deep residual neural network-based classification of loaded and unloaded UAV images

2020

Data confidence 85 high

How this score is calculated

It is built from checkable properties of the record itself: whether the article has a DOI, how many independent sources confirmed its data, whether the required fields are filled in, and how recent the latest confirmation is. Disagreement between sources and a missing identifier lower it. A value of 80 or above means the record was published without further checking; 60 to 79 means it is published at reduced confidence, with a manual check advisable. A record verified by an editor by hand is never pushed below 90 by a recalculation. This scores our data about the article, not the article itself.

Abstract

Like any new technology, unmanned aerial vehicles are used not only for good purposes. Nowadays attackers adapted UAVs for drug delivery, transportation of explosives and surveillance. For this reason, UAV detection and classification are the significant problems for researchers of this area. Previous studies in the field of UAV classification have mostly focused on classifying UAV images as UAV and no UAV, or UAV and other flying objects, also classifying different UAV models. This paper proposes a deep residual convolutional neural network based classification of loaded and unloaded UAV images. As the depth of neural network increases it shows a large learning error. In this case it is relatively easy to optimize residual neural network. Also, ResNet makes it easy to increase accuracy by increasing depth, which is more difficult to achieve with other networks. This paper attempts to show that using ResNet-34 for classification of loaded and unloaded UAV images gives superior performance and acceptable accuracy.