A New Individual Tree Species Recognition Method Based on a Convolutional Neural Network and High-Spatial Resolution Remote Sensing Imagery

Shijie Yan,Linhai Jing,Huan Wang

doi:10.3390/rs13030479

Abstract

Tree species surveys are crucial to forest resource management and can provide references for forest protection policy making. The traditional tree species survey in the field is labor-intensive and time-consuming, supporting the practical significance of remote sensing. The availability of high-resolution satellite remote sensing data enable individual tree species (ITS) recognition at low cost. In this study, the potential of the combination of such images and a convolutional neural network (CNN) to recognize ITS was explored. Firstly, individual tree crowns were delineated from a high-spatial resolution WorldView-3 (WV3) image and manually labeled as different tree species. Next, a dataset of the image subsets of the labeled individual tree crowns was built, and several CNN models were trained based on the dataset for ITS recognition. The models were then applied to the WV3 image. The results show that the distribution maps of six ITS offered an overall accuracy of 82.7% and a kappa coefficient of 0.79 based on the modified GoogLeNet, which used the multi-scale convolution kernel to extract features of the tree crown samples and was modified for small-scale samples. The ITS recognition method proposed in this study, with multi-scale individual tree crown delineation, avoids artificial tree crown delineation. Compared with the random forest (RF) and support vector machine (SVM) approaches, this method can automatically extract features and outperform RF and SVM in the classification of six tree species.

Highlights

Forests are among the most important terrestrial ecosystems and are essential for human development [1]
We explored the combination of high-resolution satellite remote sensing imagery and convolutional neural network (CNN) to recognize individual tree species (ITS)
For CNN models, GoogLeNet achieved the best overall accuracy (OA) (82.7%) with the highest kappa coefficient (0.79), and was the only model that achieved an OA over 80%

Summary

Introduction

Forests are among the most important terrestrial ecosystems and are essential for human development [1]. Well-managed forests provide renewable resources, protect biodiversity, maintain a stable energy cycle, and prevent soil degradation and erosion [2]. Traditional survey methods are inefficient and their associated labor costs are high. Remote sensing-based methods are efficient when mapping forest types in areas with rough terrain or that are difficult to reach, and can significantly improve survey efficiency and reduce labor costs [4]. Many remote sensing-based forest classification studies have considered multi-scale remote sensing data sources. Developments used medium-spatial resolution satellite remote sensing data, such as Landsat Thematic Mapper imagery, for regional-scale forest classification [5,6]. Because the spatial resolution of Landsat data is relatively low, Remote Sens.

Methods

Results

Conclusion