Hyperspectral Classification Based on 3D Convolutional Neural Network and Super Pixel Segmentation

Qiang Guo; Long Peng

doi:10.3788/AOS202141.2210001

Acta Optica Sinica, Volume. 41, Issue 22, 2210001(2021)

Hyperspectral Classification Based on 3D Convolutional Neural Network and Super Pixel Segmentation

Qiang Guo^* and Long Peng

Author Affiliations

College of Physics and Optoelectronics Engineering, Harbin Engineering University, Harbin, Heilongjiang 150001, China

show less

Abstract Get PDF(in Chinese)

Figures & Tables(9)

Fig. 1. Structure of 3D-CNN

Download full size

Fig. 2. False color image and ground truth. (a) Salinas data set; (b) Pavia University data set

Download full size

Fig. 3. PCA1 segmentation and clustering graph. (a) PCA1; (b) super pixel segmentation result; (c) merge result; (d) clustering result

Download full size

Fig. 4. PCA2 segmentation and clustering graph. (a) PCA2; (b) super pixel segmentation result; (c) merge result; (d) clustering result

Download full size

Fig. 5. PCA3 segmentation and clustering graph. (a) PCA3; (b) super pixel segmentation result; (c) merge result; (d) clustering result

Download full size

Fig. 6. PCA1 super pixel region merging results under different number of clustering centers. (a)(b) 300; (c)(d) 500; (e)(f) 1000; (g)(h) 1500; (i)(j) 2000

Download full size

Table 1. Classification accuracy of different clustering centers in Pavia University data set
View table
Table 1. Classification accuracy of different clustering centers in Pavia University data set
Parameter Center 300 Center 500 Center 1000 Center 1500 Center 2000
OA /% 86.14 92.63 97.53 97.59 96.96
AA /% 82.65 89.06 96.36 96.53 95.28
Kappa /% 83.26 91.27 96.67 95.42 93.74

Table 2. Classification accuracy of different clustering centers in Salinas data set
View table
Table 2. Classification accuracy of different clustering centers in Salinas data set
Cluster center Center 300 Center 500 Center 1000 Center 1500 Center 2000
OA /% 91.34 94.63 98.48 98.59 97.92
AA /% 90.68 93.25 97.79 97.60 97.46
Kappa /% 90.61 92.61 95.61 95.48 94.80

Table 3. Comparison of classification accuracy of different methods in Pavia University data setunit: %

View table

Table 3. Comparison of classification accuracy of different methods in Pavia University data setunit: %

Category	SVM	3D-CNN	LBP-3D-CNN	SP-3D-CNN
Asphalt	87.43	90.22	97.47	97.41
Meadows	95.80	97.95	97.11	99.65
Gravel	59.03	62.09	94.30	91.49
Trees	84.57	94.34	95.69	96.64
Metal Sheets	98.03	98.80	94.90	98.70
Soil	53.67	85.92	97.74	96.89
Bitumen	65.31	74.39	95.07	94.52
Bricks	77.59	85.58	94.13	94.10
Shadows	80.87	86.60	95.62	97.86
OA	84.16	90.29	96.19	97.53
AA	78.03	86.21	95.78	96.36
Kappa	78.57	88.39	95.70	96.67

Tools

Get Citation

Copy Citation Text

Qiang Guo, Long Peng. Hyperspectral Classification Based on 3D Convolutional Neural Network and Super Pixel Segmentation[J]. Acta Optica Sinica, 2021, 41(22): 2210001

Download Citation

EndNote(RIS)BibTex Plain Text

Set citation alerts for article

Save article for my favorites

Paper Information

Category: Image Processing

Received: Mar. 1, 2021

Accepted: May. 31, 2021

Published Online: Nov. 17, 2021

The Author Email: Guo Qiang (958542705@qq.com)

DOI:10.3788/AOS202141.2210001

Topics

laser devices and laser physics

Lasers and Laser Optics

Laser physics

laser manufacturing

Instrumentation, Measurement and Metrology

Table 1. Classification accuracy of different clustering centers in Pavia University data set

Table 1. Classification accuracy of different clustering centers in Pavia University data set

Table 2. Classification accuracy of different clustering centers in Salinas data set

Table 2. Classification accuracy of different clustering centers in Salinas data set

Table 3. Comparison of classification accuracy of different methods in Pavia University data setunit: %

Table 3. Comparison of classification accuracy of different methods in Pavia University data setunit: %