基于CNN的吉林一号卫星城市土地覆被制图潜力评估
|
吕冬梅,马玥,李华朋
|
Evaluating the Potential of JL1 Remote Sensing Data in Urban Land Cover Classification Using Convolutional Neural Networks
|
Lü Dongmei,Yue Ma,Huapeng Li
|
|
表4 S1实验区CNN模型的混淆矩阵
|
Table 4 Confusion matrix of CNN model in S1 zone
|
|
S1实验区 | 混凝土屋顶 | 金属屋顶 | 黏土屋顶 | 塑胶表面 | 沥青路面 | 铁路 | 林地 | 草地 | 裸土 | 总和 | UA/% |
---|
混凝土屋顶 | 469 | 12 | 18 | 0 | 53 | 12 | 20 | 0 | 13 | 597 | 78.56 | 金属屋顶 | 0 | 488 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 488 | 100 | 黏土屋顶 | 12 | 0 | 328 | 3 | 0 | 0 | 0 | 0 | 0 | 343 | 95.63 | 塑胶表面 | 0 | 0 | 1 | 241 | 17 | 0 | 0 | 0 | 25 | 284 | 84.86 | 沥青路面 | 26 | 1 | 0 | 1 | 335 | 2 | 0 | 1 | 27 | 393 | 85.24 | 铁路 | 16 | 0 | 12 | 0 | 19 | 194 | 0 | 0 | 14 | 255 | 76.08 | 林地 | 0 | 0 | 0 | 0 | 3 | 2 | 493 | 39 | 6 | 543 | 90.79 | 草地 | 0 | 0 | 0 | 23 | 2 | 0 | 0 | 263 | 0 | 288 | 91.32 | 裸土 | 18 | 0 | 12 | 41 | 9 | 11 | 0 | 0 | 224 | 315 | 71.11 | 总和 | 541 | 501 | 371 | 309 | 438 | 221 | 513 | 303 | 309 | 3 506 | | PA/% | 86.69 | 97.41 | 88.41 | 77.99 | 76.48 | 87.78 | 96.1 | 86.8 | 72.49 | | | OA/% | 86.57 | | | Kappa | 0.847 4 | | |
|
|
|