TEL:886-2-2367-8532
FAX:886-2-2368-6732

Full-text available
   < Previous vol/No | Next vol/No >
Vol.51, No.4, PP.127-172
Content:   PDF( 0.64MB )

Editorial Scope and Editorial Board:  PDF( 0.28MB )

1
Deep Neural Networks for Object-Based Image Classification of Hyperspectral Images
51(4):127-139
Ming-Der Yang[1,2] Hsin-Hung Tseng[1,2]* Yi-Chin Hung[1] Yu-Chun Hsu[1,2]
* Corresponding Author. E-mail : d108062001@mail.nchu.edu.tw
Show preview   |   PDF( 6.4MB )   |  
Deep Neural Networks for Object-Based Image Classification of Hyperspectral Images
Close
Ming-Der Yang[1,2] Hsin-Hung Tseng[1,2]* Yi-Chin Hung[1] Yu-Chun Hsu[1,2]

Abstract
In hyperspectral images, each pixel contains information from hundreds of frequency bands, which is helpful for improving the classification accuracy; however, this also reduces the processing rate. In conventional pixelbased
image classification, pixel data rarely represent spatial correlations between objects. Thus, an object-based deep
neural network (DNN) was developed to classify hyperspectral images.
The research data included images of Indian Pines (Indiana, USA) and Salinas (California, USA) recorded by an airborne
visible/infrared imaging spectrometer, along with the associated ground truths for image classification. First, the minimum noise fraction technique was applied to separate the noise from the images in order to provide useful information and reduce the need of calculations. Then, object-based image analysis was implemented to explore spatial correlation. Simple linear iterative clustering was then used to categorize the objects based on size and compactness of each object. Finally, applying DNN classification can solve the salt-and-pepper effect, and requires less computing time than pixel-based classification, especially for a large area. The classification accuracy and kappa of the objectbased
classification were higher than those of the pixel-based classification method. The proposed method was verified
using an image of Salinas with a classification accuracy of 94.62 % and less computation time.
Key Words: hyperspectral image, minimum noise fraction, simple linear iterative clustering, deep neural network
〔1〕Department of Civil Engineering, and Innovation and Development Center of Sustainable Agriculture, National Chung Hsing University,Taichung 402, Taiwan, R.O.C.
〔2〕Pervasive AI Research (PAIR) Labs, Taiwan, R.O.C.
* Corresponding Author. E-mail : d108062001@mail.nchu.edu.tw
Received: 2020/02/17
Revised: 2020/06/08
Accepted: 2021/01/08
2
Applying Portable Magnetometers to Detect Buried Instruments After a Debris Flow Event
51(4):140-146
Hsien-Te Chou[1*] Chih-Hsuan Huang[1] Horng -Yuan Yen[2] Tao-Ssu Tsai[3]
* Corresponding Author. E-mail : htchou@cc.ncu.edu.tw
Show preview   |   PDF( 7.05MB )   |  
3
Long-Term Geomorphologic and Landslide Evolution in the Heshe River Watershed After Frequent Sediment Disasters
51(4):147-158
Chun-Hung Wu Zhou-Ting Zhou* Cheng-Yi Lin
* Corresponding Author. E-mail : wdrg1122@gmail.com
Show preview   |   PDF( 7.25MB )   |  
4
Planning the Extent of Fluvial Corridors in Taiwan
51(4):159-168
Chia-Ning Yang Cheng-Wei Kuo*
* Corresponding Author. E-mail : cwkuo@mail.sinotech.com.tw
Show preview   |   PDF( 12.07MB )   |  
   Copyright © 2011 Chinese Soil And Water Conservation Society   MAP    
Address:2FL, No.8, Lane 113, Xiamen St., Zhongzheng Dist., Taipei City 100, Taiwan (R.O.C.)