Identifying the vegetation type in Google Earth images using a convolutional neural network: a case study for Japanese bamboo forests

Kazuaki Sumi,Takeshi Ise,Shuntaro Watanabe

doi:10.1186/s12898-020-00331-5

Kazuaki Sumi, Takeshi Ise

Open Access

https://doi.org/10.1186/s12898-020-00331-5

Copy DOI

Journal: BMC Ecology	Publication Date: Nov 27, 2020
Citations: 16	License type: open-access

Affiliation: Kyoto University, Kagoshima University

Abstract

BackgroundClassifying and mapping vegetation are crucial tasks in environmental science and natural resource management. However, these tasks are difficult because conventional methods such as field surveys are highly labor-intensive. Identification of target objects from visual data using computer techniques is one of the most promising techniques to reduce the costs and labor for vegetation mapping. Although deep learning and convolutional neural networks (CNNs) have become a new solution for image recognition and classification recently, in general, detection of ambiguous objects such as vegetation is still difficult. In this study, we investigated the effectiveness of adopting the chopped picture method, a recently described protocol for CNNs, and evaluated the efficiency of CNN for plant community detection from Google Earth images.ResultsWe selected bamboo forests as the target and obtained Google Earth images from three regions in Japan. By applying CNN, the best trained model correctly detected over 90% of the targets. Our results showed that the identification accuracy of CNN is higher than that of conventional machine learning methods.ConclusionsOur results demonstrated that CNN and the chopped picture method are potentially powerful tools for high-accuracy automated detection and mapping of vegetation.

Highlights

Classifying and mapping vegetation are crucial tasks in environmental science and natural resource management
We investigated the efficiency of adopting a deep learning model and the chopped picture method for computer-based vegetation detection from Google Earth images
Comparison with support vector machine (SVM) Our results show that the convolutional neural network (CNN) classifier outperformed the SVM classifier (Fig. 7)

Summary

Introduction

Classifying and mapping vegetation are crucial tasks in environmental science and natural resource management. Deep learning and convolutional neural networks (CNNs) have become a new solution for image recognition and classification recently, in general, detection of ambiguous objects such as vegetation is still difficult. Remote sensing technology offers a practical and economical means to acquire information on Detection of discriminating visual features is one of the most important steps in almost all computer vision problems, including in the field of remote sensing. Because conventional methods such as support vector machines [3] require hand-designed, time-consuming feature extraction, substantial efforts have been dedicated toward the development of methods for the automatic extraction of features. Recent research has shown that the CNN can successfully detect plant diseases and accurately classify plant specimens in an herbarium [8,9,10]

Methods

Results

Conclusion