Tree Species Classification Based on Hybrid Ensembles of a Convolutional Neural Network (CNN) and Random Forest Classifiers

被引：38

作者：

Knauer, Uwe ^{[1
]}

von Rekowski, Cornelius Styp ^{[2
]}

Stecklina, Marianne ^{[2
]}

Krokotsch, Tilman ^{[2
]}

Tuan Pham Minh ^{[2
]}

Hauffe, Viola ^{[2
]}

Kilias, David ^{[1
]}

Ehrhardt, Ina ^{[1
]}

Sagischewski, Herbert ^{[3
]}

Chmara, Sergej ^{[3
]}

Seiffert, Udo ^{[1
]}

机构：

[1] Fraunhofer Inst Factory Operat & Automat, D-39106 Magdeburg, Germany

[2] Otto von Guericke Univ, Fac Comp Sci, D-39106 Magdeburg, Germany

[3] Forstl Forsch & Kompetenzzentrum, ThuringenForst AoR, D-99867 Gotha, Germany

来源：

REMOTE SENSING | 2019年 / 11卷 / 23期

关键词：

hyperspectral imaging; tree species; multiple classifier fusion; convolutional neural network; random forest; rotation forest; HYPERSPECTRAL DATA; REFLECTANCE; SVM;

D O I：

10.3390/rs11232788

中图分类号：

X [环境科学、安全科学];

学科分类号：

08 ; 0830 ;

摘要：

In this paper, we evaluate different popular voting strategies for fusion of classifier results. A convolutional neural network (CNN) and different variants of random forest (RF) classifiers were trained to discriminate between 15 tree species based on airborne hyperspectral imaging data. The spectral data was preprocessed with a multi-class linear discriminant analysis (MCLDA) as a means to reduce dimensionality and to obtain spatial-spectral features. The best individual classifier was a CNN with a classification accuracy of 0.73 +/- 0.086. The classification performance increased to an accuracy of 0.78 +/- 0.053 by using precision weighted voting for a hybrid ensemble of the CNN and two RF classifiers. This voting strategy clearly outperformed majority voting (0.74), accuracy weighted voting (0.75), and presidential voting (0.75).

引用

页数：15

共 36 条

[1]

[Anonymous], 2013, STAT LEARNING THEORY

[2] Spatially regularized semisupervised Ensembles of Extreme Learning Machines for hyperspectral image segmentation [J].