基于双树复小波分解的Boosting集成学习土地覆被分类研究
|
李润祥,高小红,汤敏
|
Study on Boosting Ensemble Learning Land Cover Classification based on Dual-Tree Complex Wavelet Transform
|
Runxiang Li,Xiaohong Gao,Min Tang
|
|
表3 Sentinel-2A影像分类精度
|
Table 3 Classification accuracy of Sentinel-2A image
|
|
土地覆被类型 | RF | GBDT | DTCWT-GBDT | XGBoost | DTCWT-XGBoost | LightGBM | DTCWT-LightGBM |
---|
用户精度/% | 制图精度/% | 用户精度/% | 制图精度/% | 用户精度/% | 制图精度/% | 用户精度/% | 制图精度/% | 用户精度/% | 制图精度/% | 用户精度/% | 制图精度/% | 用户精度/% | 制图精度/% |
---|
耕地 | 82.63 | 81.21 | 85.06 | 83.26 | 86.63 | 83.62 | 87.75 | 86.32 | 88.84 | 86.98 | 89.64 | 89.01 | 90.96 | 86.68 | 有林地 | 91.54 | 90.12 | 84.50 | 80.23 | 90.51 | 92.32 | 92.63 | 92.76 | 93.69 | 92.45 | 93.68 | 93.86 | 93.84 | 92.69 | 灌木林地 | 86.23 | 88.23 | 87.36 | 88.26 | 87.54 | 89.31 | 88.23 | 89.36 | 88.36 | 89.35 | 89.45 | 89.23 | 89.85 | 90.21 | 疏林地、其他林地 | 90.12 | 93.63 | 90.15 | 88.45 | 92.35 | 91.94 | 93.32 | 91.88 | 92.88 | 93.45 | 94.63 | 95.54 | 96.56 | 95.34 | 高覆盖草地 | 84.26 | 90.23 | 84.63 | 85.66 | 84.76 | 84.55 | 85.46 | 83.52 | 86.65 | 84.78 | 86.81 | 83.65 | 87.82 | 86.45 | 中覆盖草地 | 88.23 | 80.52 | 90.36 | 82.77 | 93.42 | 85.63 | 91.38 | 86.23 | 92.66 | 91.56 | 93.36 | 91.39 | 95.35 | 92.36 | 低覆盖草地 | 83.12 | 84.33 | 85.42 | 81.56 | 86.06 | 85.43 | 86.82 | 84.67 | 87.26 | 86.45 | 87.82 | 84.82 | 88.36 | 85.42 | 城乡工矿居住 | 90.29 | 85.36 | 91.65 | 92.34 | 91.78 | 91.23 | 91.93 | 91.62 | 92.16 | 93.45 | 93.35 | 94.26 | 94.42 | 95.32 | 建设用地 | 河流 | 94.75 | 95.26 | 95.31 | 96.36 | 96.56 | 96.23 | 96.76 | 96.67 | 96.53 | 96.48 | 95.69 | 96.26 | 96.59 | 95.36 | 水库坑塘 | 96.11 | 93.16 | 97.46 | 93.23 | 97.32 | 95.13 | 97.15 | 97.23 | 98.36 | 95.06 | 98.33 | 99.14 | 99.21 | 98.36 | 未利用土地 | 80.12 | 80.07 | 83.53 | 81.26 | 84.21 | 82.14 | 84.68 | 83.47 | 87.93 | 82.63 | 88.45 | 85.36 | 89.18 | 87.66 | 总体精度/% | 87.85 | 89.19 | 91.23 | 90.58 | 92.46 | 91.72 | 93.25 | Kappa | 0.86 | 0.87 | 0.90 | 0.88 | 0.89 | 0.90 | 0.91 |
|
|
|