I am using Anaconda (Python 2.7) and the GDAL package from OSGEO.
- Choose the number of trees for the Random Forest (default = 500)
- The multi-band Remote Sensing image as tif format.
- An image which defines your training dataset in the same extend as the remote sensing image
- An image which defines your validation dataset in the same extend as the remote sensing image
In new version "*_shape.py" training and validation data can be inputted directly as shape files!
Point 3 and 4 are noit necessary anymore!
Furthermore, the classification result will be masked. Thus a black border has a value of 0 in the classification.
This is the only section where you have to change something in the script (directories and file names).
This example uses a 14 bands remote sensing dataset and 11 classes as training and validation. In the end, a Tiff-File will be saved as your classification. During the process you will see 2 plots and a lot of information...
We have 87895 training samples
The training data include 11 classes: [ 1 2 3 4 5 6 7 8 9 10 11]
Our X matrix is sized: (87895L, 14L)
Our y array is sized: (87895L,)
Our OOB prediction of accuracy is: 99.8634734627%
Band 1 importance: 0.0261680458145
Band 2 importance: 0.0153287127996
Band 3 importance: 0.0205070224479
Band 4 importance: 0.037036238847
Band 5 importance: 0.0310529483133
Band 6 importance: 0.0436272835319
Band 7 importance: 0.055942106673
Band 8 importance: 0.398011332962
Band 9 importance: 0.119100238254
Band 10 importance: 0.0119935702383
Band 11 importance: 0.0857394486887
Band 12 importance: 0.0216590258319
Band 13 importance: 0.0998573985857
Band 14 importance: 0.033976627012
predict 1 2 3 4 5 6 7 8 9 10 11 All
truth
1 5280 0 0 0 0 0 0 0 0 0 0 5280
2 0 170 0 0 0 0 0 0 0 0 0 170
3 0 0 2962 0 0 0 0 0 0 0 0 2962
4 0 0 0 16650 0 0 0 0 0 0 0 16650
5 0 0 0 0 6194 0 0 0 0 0 0 6194
6 0 0 0 0 0 1691 0 0 0 0 0 1691
7 0 0 0 0 0 0 4509 0 0 0 0 4509
8 0 0 0 0 0 0 0 2342 0 0 0 2342
9 0 0 0 0 0 0 0 0 41504 0 0 41504
10 0 0 0 0 0 0 0 0 0 6218 0 6218
11 0 0 0 0 0 0 0 0 0 0 375 375
All 5280 170 2962 16650 6194 1691 4509 2342 41504 6218 375 87895
We have 68482 validation pixels
The validation data include 11 classes: [ 1 2 3 4 5 6 7 8 9 10 11]
Our X matrix is sized: (68482L,)
Our y array is sized: (68482L,)
col_0 1 2 3 4 5 6 7 8 9 10 11 All
row_0
1 2597 0 0 0 0 0 0 0 701 0 0 3298
2 0 125 0 0 0 0 0 0 0 0 10 135
3 0 0 1457 743 0 0 0 16 0 85 0 2301
4 0 0 163 9640 12 0 0 473 0 72 0 10360
5 2 0 0 410 5505 0 0 0 858 47 0 6822
6 0 0 0 128 0 934 0 7 0 483 0 1552
7 0 0 0 0 0 0 2662 0 0 0 0 2662
8 0 0 85 389 8 0 0 538 0 299 0 1319
9 137 0 0 3 1620 0 0 0 31385 0 0 33145
10 0 0 47 520 63 190 40 19 1 5656 0 6536
11 0 0 0 0 0 0 0 0 0 0 352 352
All 2736 125 1752 11833 7208 1124 2702 1053 32945 6642 362 68482
precision recall f1-score support
1 0.95 0.79 0.86 3298
2 1.00 0.93 0.96 135
3 0.83 0.63 0.72 2301
4 0.81 0.93 0.87 10360
5 0.76 0.81 0.78 6822
6 0.83 0.60 0.70 1552
7 0.99 1.00 0.99 2662
8 0.51 0.41 0.45 1319
9 0.95 0.95 0.95 33145
10 0.85 0.87 0.86 6536
11 0.97 1.00 0.99 352
avg / total 0.89 0.89 0.89 68482
OAA = 88.8569259075 %