Landsat unsupervised classification Zhuosen Wang 1
Feb 20, 2016
1
Landsat unsupervised classification
Zhuosen Wang
2
Unsupervised classification methods
• The two most frequently used algorithms K-mean and the ISODATA
• Minimize the distance between each pixel and its assigned cluster center
• The ISODATA algorithm allows for different number of clusters while the k-means assumes that the number of clusters is known a priori
• K-means is very sensitive to initial starting values
3
15 classes , 1 iteration 7 classes, 5 iterations K-mean P028r035
4
7 classes 5 iteration IsoDATA K-mean P028r035
5
7 classes 5 iteration IsoDATA K-mean
P028r035
6
10 classes 7 classes 5 iterations, IsoDATA
7
7 classes, 5 iterations IsoDataP12r31 –2011_09_02
8
7 classes, 5 iterations 7 classes, 10 iterations IsoData P12r31 –2011_09_02 No improvement between 10 iterations and 5 iterationsCyan –grass Yellow –deciduous forest blue,green—evergreen forest
K-mean IsoDATA 7 classes, 5 iterations P12r31 –2011_09_02
10
Harvard Forest
11
Harvard Forest p012r030