Forest Species Classification of UAV Hyperspectral Image Using Deep Learning
| Autor: | Hui Zhao, Jing Liang, Mingliang Qu, Pengshuai Li, Lu Han |
|---|---|
| Rok vydání: | 2020 |
| Předmět: |
Pixel
Computer science business.industry Deep learning 010401 analytical chemistry Feature extraction Hyperspectral imaging Pattern recognition 02 engineering and technology Vegetation 021001 nanoscience & nanotechnology 01 natural sciences Convolutional neural network 0104 chemical sciences Image (mathematics) Softmax function Artificial intelligence 0210 nano-technology business |
| Zdroj: | 2020 Chinese Automation Congress (CAC). |
| DOI: | 10.1109/cac51589.2020.9327690 |
| Popis: | Forest species classification is essential for surveying of forest resource, biodiversity research, and community structure. The tree species level classification can be achieved by hyperspectral image(HSI), since the HSI has the high resolution of spectral and spatial.Classifying and mapping the forest species by HSI can be converted to classify each pixel vector of HSI. In this paper, we propose a spectral–spatial paralleled convolutional neural network(SSPCNN) to classify the forest tree species in the UAV(unmanned aerial vehicle) HSI. The SSPCNN mainly consists of one-dimensional convolutional neural network(1-D-CNN) and two-dimensional convolutional neural network(2-D-CNN). The 1-D-CNN is used to learn the spectral features , and the 2-D-CNN is applied to extract the spatial features. Finally two types features are fused to classify by the softmax classifier.The experimental result shows that the SSPCNN produced competitive performance compared with other methods. |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|