Multiple object extraction from aerial imagery with convolutional neural networks

Shunta Saito, Takayoshi Yamashita, Yoshimitsu Aoki

Research output: Contribution to journalConference article

24 Citations (Scopus)

Abstract

An automatic system to extract terrestrial objects from aerial imagery has many applications in a wide range of areas. However, in general, this task has been performed by human experts manually, so that it is very costly and time consuming. There have been many attempts at automating this task, but many of the existing works are based on class-specific features and classifiers. In this article, the authors propose a convolutional neural network (CNN)-based building and road extraction system. This takes raw pixel values in aerial imagery as input and outputs predicted three-channel label images (building-road-background). Using CNNs, both feature extractors and classifiers are automatically constructed. The authors propose a new technique to train a single CNN efficiently for extracting multiple kinds of objects simultaneously. Finally, they show that the proposed technique improves the prediction performance and surpasses state-of-the-art results tested on a publicly available aerial imagery dataset.

Original languageEnglish
JournalIS and T International Symposium on Electronic Imaging Science and Technology
DOIs
Publication statusPublished - 2016 Jan 1
Event33rd Intelligent Robots and Computer Vision: Algorithms and Techniques Conference - San Francisco, United States
Duration: 2016 Feb 142016 Feb 18

Fingerprint

aerial photography
Antennas
classifiers
Neural networks
roads
Classifiers
performance prediction
Labels
Pixels
pixels
output

ASJC Scopus subject areas

  • Computer Graphics and Computer-Aided Design
  • Computer Science Applications
  • Human-Computer Interaction
  • Software
  • Electrical and Electronic Engineering
  • Atomic and Molecular Physics, and Optics

Cite this

Multiple object extraction from aerial imagery with convolutional neural networks. / Saito, Shunta; Yamashita, Takayoshi; Aoki, Yoshimitsu.

In: IS and T International Symposium on Electronic Imaging Science and Technology, 01.01.2016.

Research output: Contribution to journalConference article

@article{dd41c8400115402cb080e1397e35d7d7,
title = "Multiple object extraction from aerial imagery with convolutional neural networks",
abstract = "An automatic system to extract terrestrial objects from aerial imagery has many applications in a wide range of areas. However, in general, this task has been performed by human experts manually, so that it is very costly and time consuming. There have been many attempts at automating this task, but many of the existing works are based on class-specific features and classifiers. In this article, the authors propose a convolutional neural network (CNN)-based building and road extraction system. This takes raw pixel values in aerial imagery as input and outputs predicted three-channel label images (building-road-background). Using CNNs, both feature extractors and classifiers are automatically constructed. The authors propose a new technique to train a single CNN efficiently for extracting multiple kinds of objects simultaneously. Finally, they show that the proposed technique improves the prediction performance and surpasses state-of-the-art results tested on a publicly available aerial imagery dataset.",
author = "Shunta Saito and Takayoshi Yamashita and Yoshimitsu Aoki",
year = "2016",
month = "1",
day = "1",
doi = "10.2352/ISSN.2470-1173.2016.10.ROBVIS-392",
language = "English",
journal = "IS and T International Symposium on Electronic Imaging Science and Technology",
issn = "2470-1173",

}

TY - JOUR

T1 - Multiple object extraction from aerial imagery with convolutional neural networks

AU - Saito, Shunta

AU - Yamashita, Takayoshi

AU - Aoki, Yoshimitsu

PY - 2016/1/1

Y1 - 2016/1/1

N2 - An automatic system to extract terrestrial objects from aerial imagery has many applications in a wide range of areas. However, in general, this task has been performed by human experts manually, so that it is very costly and time consuming. There have been many attempts at automating this task, but many of the existing works are based on class-specific features and classifiers. In this article, the authors propose a convolutional neural network (CNN)-based building and road extraction system. This takes raw pixel values in aerial imagery as input and outputs predicted three-channel label images (building-road-background). Using CNNs, both feature extractors and classifiers are automatically constructed. The authors propose a new technique to train a single CNN efficiently for extracting multiple kinds of objects simultaneously. Finally, they show that the proposed technique improves the prediction performance and surpasses state-of-the-art results tested on a publicly available aerial imagery dataset.

AB - An automatic system to extract terrestrial objects from aerial imagery has many applications in a wide range of areas. However, in general, this task has been performed by human experts manually, so that it is very costly and time consuming. There have been many attempts at automating this task, but many of the existing works are based on class-specific features and classifiers. In this article, the authors propose a convolutional neural network (CNN)-based building and road extraction system. This takes raw pixel values in aerial imagery as input and outputs predicted three-channel label images (building-road-background). Using CNNs, both feature extractors and classifiers are automatically constructed. The authors propose a new technique to train a single CNN efficiently for extracting multiple kinds of objects simultaneously. Finally, they show that the proposed technique improves the prediction performance and surpasses state-of-the-art results tested on a publicly available aerial imagery dataset.

UR - http://www.scopus.com/inward/record.url?scp=85046053800&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85046053800&partnerID=8YFLogxK

U2 - 10.2352/ISSN.2470-1173.2016.10.ROBVIS-392

DO - 10.2352/ISSN.2470-1173.2016.10.ROBVIS-392

M3 - Conference article

AN - SCOPUS:85046053800

JO - IS and T International Symposium on Electronic Imaging Science and Technology

JF - IS and T International Symposium on Electronic Imaging Science and Technology

SN - 2470-1173

ER -