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5 S2实验区CNN模型的混淆矩阵
|
Table 5 Confusion matrix of CNN model in S2 zone
|
|
S2实验区 | 混凝土屋顶 | 金属屋顶 | 黏土屋顶 | 塑胶表面 | 沥青路面 | 林地 | 草地 | 裸土 | 水体 | 总数 | UA/% |
---|
混凝土屋顶 | 454 | 16 | 5 | 9 | 19 | 0 | 0 | 0 | 0 | 503 | 90.26 | 金属屋顶 | 0 | 336 | 0 | 0 | 17 | 0 | 0 | 0 | 0 | 353 | 95.18 | 黏土屋顶 | 4 | 0 | 470 | 13 | 0 | 0 | 0 | 0 | 0 | 487 | 96.51 | 塑胶表面 | 8 | 0 | 10 | 257 | 2 | 0 | 0 | 0 | 0 | 277 | 92.78 | 沥青路面 | 17 | 0 | 0 | 12 | 384 | 0 | 1 | 0 | 0 | 414 | 92.75 | 林地 | 0 | 0 | 0 | 0 | 16 | 444 | 9 | 0 | 22 | 491 | 90.43 | 草地 | 0 | 0 | 0 | 0 | 1 | 3 | 295 | 27 | 0 | 326 | 90.49 | 裸土 | 15 | 0 | 0 | 2 | 6 | 0 | 6 | 274 | 0 | 303 | 90.43 | 水体 | 9 | 0 | 0 | 0 | 38 | 11 | 0 | 0 | 282 | 340 | 82.94 | 总数 | 507 | 352 | 485 | 293 | 483 | 458 | 311 | 301 | 304 | 3 494 | | PA/% | 89.55 | 95.45 | 96.91 | 87.71 | 79.50 | 96.94 | 94.86 | 91.03 | 92.76 | | | OA/% | 91.47 | | | Kappa | 0.903 5 | | |
|
|
|