Hyperspectral remote sensing image classification based on dense residual three-dimensional convolutional neural network
- PDF / 2,724,150 Bytes
- 24 Pages / 439.37 x 666.142 pts Page_size
- 45 Downloads / 237 Views
Hyperspectral remote sensing image classification based on dense residual three-dimensional convolutional neural network Suting Chen 1,2
1
& Meng Jin & Jie Ding
2
Received: 30 July 2019 / Revised: 22 June 2020 / Accepted: 28 July 2020 # Springer Science+Business Media, LLC, part of Springer Nature 2020
Abstract
Data-driven deep learning techniques set the current state of the art in image classification for hyperspectral remote sensing images. The lack of labeled training data and high dimensionality of hyperspectral images, results in these techniques being far from satisfactory with respect to accuracy and efficiency. To address the deficiencies of the existing approaches, we proposed a novel neural network technique, namely, dense residual three-dimensional convolutional neural network (DR3D-CNN). Tailored for hyperspectral images, this network used 3D convolution instead of the conventional 2D convolution for more effective spectral feature extraction. It also employed dense residual connections to alleviate the problem of gradient dispersion. After the initial classification by the network, the proposed technique further refined the result using multi-label conditional random field optimization. Experimental results on various hyperspectral image datasets showed that the proposed model outperforms existing deep learning techniques with respect to accuracy by a large margin while requiring less training time. Keywords Hyperspectral remote sensing classification . Deep convolution . Three-dimensional convolution . Dense residual connection . Multi-label conditional random field
* Suting Chen [email protected] Meng Jin [email protected] Jie Ding [email protected]
1
Jiangsu Key Laboratory of Meteorological Observation and Information Processing, Nanjing University of Information Science & Technology, Nanjing 210044, China
2
Jiangsu Collaborative Innovation Center of Atmospheric Environment and Equipment Technology (CICAEET), Nanjing University of Information Science & Technology, Nanjing 210044, China
Multimedia Tools and Applications
1 Introduction With the development of hyperspectral remote sensing technology, the new hyperspectral sensor is capable of simultaneously acquiring continuous images of spectral features and spatial features, which contains rich feature information. Remote sensing data not only reflects the spectral information of land cover but also includes the spatial distribution information of land cover. Consequently, it has a wide range of applications in agriculture, environmental monitoring, urban planning [14], and military surveillance [41]. The development and changes of land cover cause changes in the values of remote sensing images. By analyzing the variation of remote sensing image data, the types of land cover can be effectively classified and identified. The use of hyperspectral data for land cover classification and object recognition is a research hotspot in hyperspectral-data processing. Domestic and foreign scholars have carried out a series of extensive an
Data Loading...
