Open AccessSemantic segmentation network of uav image based on improved U-net
Author(s) -
Ziyi Liu,
Jin Huang
Publication year - 2019
Publication title -
iop conference series. earth and environmental science
Language(s) - English
Resource type - Journals
eISSN - 1755-1307
pISSN - 1755-1315
DOI - 10.1088/1755-1315/330/5/052050
Subject(s) - net (polyhedron) , computer science , artificial intelligence , segmentation , field (mathematics) , pixel , convolutional neural network , sample (material) , image (mathematics) , artificial neural network , set (abstract data type) , image segmentation , pattern recognition (psychology) , deep learning , computer vision , mathematics , chemistry , geometry , chromatography , pure mathematics , programming language
The u-net, which is popular in the field of deep learning, is improved on the basis of the Fully Convolutional Network (FCN) [1]. U-net [2] was first published in the field of medical image, because the network has many advantages, so many scholars put their migration to use on the other computer vision problems, finding its on different problems still show the excellent performance. In this paper, U-net is applied to the problem of uav image processing, and the semantic segmentation of uav images at pixel level is carried out to identify the types of ground objects and the pixels contained in them. In order to increase the recognition accuracy of neural network, this paper proposes an improved network model based on u-net, and applies it to this problem. By training, verifying and testing the same small sample data set, it is found that the improved network has better recognition accuracy than U-net.