改进3D-Octave卷积的高光谱图像分类方法

郑宗生, 王政翰, 王振华, 卢鹏, 高萌, 霍志俊

An improved 3D Octave convolution-based method for hyperspectral image classification

ZHENG Zongsheng, WANG Zhenghan, WANG Zhenhua, LU Peng, GAO Meng, HUO Zhijun

表2 不同算法在PU数据集的分类结果

Tab.2 Classification resultsof different algorithms on PU dataset(%)

类别 SVM 3D-CNN SSRN HybridSN DBDA 本文方法 1 93.37 96.40 98.25 97.26 98.80 99.94 2 94.73 99.13 99.85 99.87 99.90 100.00 3 74.46 91.14 93.05 99.09 97.53 99.50 4 90.37 97.50 97.00 99.89 99.93 98.10 5 99.92 99.24 99.92 99.23 98.17 99.77 6 89.20 99.30 96.61 99.78 99.81 99.67 7 81.68 92.64 98.58 100.00 99.61 99.60 8 79.04 90.23 92.65 96.41 98.02 98.50 9 100.00 94.22 96.95 99.77 99.88 99.51 OA 91.85 97.41 98.75 99.08 99.37 99.61 Kappa 89.22 96.56 98.34 98.68 99.17 99.44 AA 89.77 95.44 97.60 97.87 98.53 99.08