Lianchen Zhao, Yizhun Peng, Di Li, Yuheng Zhang
Page 351-356
Abstract
There are still some improvements in target detection of high-resolution optical remote sensing images. In this paper, deep learning algorithm is used to detect the target in remote sensing image. Improve and optimize the deep learning target detection algorithm. The concepts of multi-scale feature fusion, feature pyramid and residual network are explained. By improving yolov3 algorithm, the detection effect of aircraft and ship in DOTA data set is significantly improved.
Keywords: Deep learning; Target detection; Residual network; Remote sensing image; Feature pyramid