Multi-scale Dilated Convolutional Neural Network for Object Detection in UAV Images

As a new type of remote sensing sensor, unmanned aerial vehicle(UAV)has been used in various fields such as medical treatment, transportation, environmental monitoring, disaster warning, animal protection and military increasingly. Since UAV images are acquired from multiple flying altitudes, perspectives with high speed, objects in UAV images have various scales and perspectives with different distributions, which brings a series of problems to object detection in UAV images.To address these problems, we propose an object detection method based on multi-scale dilated convolutional neural network. It improves existing detection methods by a creative multi-scale dilated convolutional module which facilitates the whole network to learn deep features with increased field of view perception and further improves the performance of object detection in UAV images.We adopt three comparative experiments on base network and our proposed method. And experimental results show that our proposed network has a high precision and recall for object detection in UAV images. Moreover, objects are detected with high performance in multiple perspectives, various scales and complex backgrounds, which indicates the effectiveness and robustness of our method.Object detection in UAV image is significant in both civil and military fields. However, existing methods are limited with objects in multiple perspectives, scales and backgrounds.Our proposed method improves the performance of existing networks by dilated convolutional operator. Experimental results demonstrate the effectiveness and robustness of the proposed method. © 2020, Editorial Board of Geomatics and Information Science of Wuhan University. All right reserved.