User Tools

Site Tools


wiki:contest_data_fusion

Differences

This shows you the differences between two versions of the page.

Link to this comparison view

wiki:contest_data_fusion [2017/12/05 22:53] (current)
Line 1: Line 1:
 +=== Stack bands in feature fusion approach ===
 +
 +<​note>​
 +LiDAR=band 0 \\
 +NDVI=band 1\\
 +LINEAR FEATURES=band 2 and 3\\
 +CASI1-144=band 4 ... band 147\\
 +</​note>​
 +
 +<code bash>
 +    pkcrop -i 2013_IEEE_GRSS_DF_Contest_LiDAR.tif -i 2013_IEEE_GRSS_DF_Contest_NDVI.tif -i 2013_IEEE_GRSS_DF_Contest_LINEAR.tif -i 2013_IEEE_GRSS_DF_Contest_CASI.tif -o 2013_IEEE_GRSS_DF_Contest_FEATURE_FUSION.tif
 +</​code>​
 +
 +==== Feature fusion ====
 +
 +=== Create training vector with stacked features ===
 +<code bash>
 +    pkextract -i 2013_IEEE_GRSS_DF_Contest_FEATURE_FUSION.tif -s 2013_IEEE_GRSS_DF_Contest_Samples_TR_26915.shp -o training_feature_fusion.shp
 +</​code>​
 +
 +=== Optimize SVM parameters for these features ===
 +<code bash>
 +    pkopt_svm -t training_feature_fusion.shp -cc 0.1 -cc 10000 -g 0.001 -g 10 -step 10
 +</​code>​
 +
 +<note tip>
 +--ccost 7657.41 --gamma 0.186167
 +</​note>​
 +
 +=== Redo the feature selection based on this new feature set (we can stop already after 10 features) ===
 +<code bash>
 +    pkfs_svm -t training_feature_fusion.shp -cc 100 -g 1 -n 10 -v 1 -cv 2
 +</​code>​
 +
 +<note tip>
 +-b 140 -b 2 -b 48 -b 0 -b 3 -b 23
 +</​note>​
 +
 +==== Image classification ====
 +
 +=== Based on the optimal feature set, we can create a new land cover map ===
 +<code bash>
 +    pkclassify_svm -i 2013_IEEE_GRSS_DF_Contest_FEATURE_FUSION.tif -t training_feature_fusion.shp -o testmap_feature_fusion.tif -ct ct.txt --ccost 100 --gamma 1 -b 140 -b 2 -b 48 -b 0 -b 3 -b 23
 +</​code>​
 +
  
wiki/contest_data_fusion.txt ยท Last modified: 2017/12/05 22:53 (external edit)