Text
KLASIFIKASI CITRA HIPERSPEKTRAL PADA TUTUPAN LAHAN MENGGUNAKAN 3D CONVOLUTIONAL NEURAL NETWORK
Land cover with hyperspectral satellite imagery can provide accurate information in the earth surface monitoring activities. Various studies have been developed on the land cover hyperspectral imagery classification, but limited public data is an obstacle in determining the right classification model. Several studies validate model performance by testing with large test data and very limited training samples and extreme data imbalance. However, these studies have complicated model structures and suboptimal performance. This study uses a method with a simpler structure and shallower network, by combining HybridSN, 3D dilated convolution and MSR3DCNN which were tested on two datasets Indian Pines (IP) and Salinas (SA). The test results showed that the combination of HybridSN and 3D dilated convolution obtained the highest accuracy compared to the HybridSN�MSR3DCNN combination as well as other studies with the same number of test samples, with OA accuracy of 96.58%, kappa 96.09% on the IP dataset and AA accuracy of 99.08%, kappa 99.00% on the SA dataset.
Inventory Code | Barcode | Call Number | Location | Status |
---|---|---|---|---|
2207004113 | T79753 | T797532022 | Central Library (Referens) | Available but not for loan - Not for Loan |
No other version available