Raster analysis with machine learning

20 Variables

I included the following variables iteratively.

  • Slope
  • Roughness
  • TPI, TMI
  • 9 moving focal windows
  • Gaussian window with a Laplacian filter

Results of the model

## Ranger result
## 
## Call:
##  ranger(formula = as.factor(class) ~ ., data = dat_train) 
## 
## Type:                             Classification 
## Number of trees:                  500 
## Sample size:                      3708 
## Number of independent variables:  20 
## Mtry:                             4 
## Target node size:                 1 
## Variable importance mode:         none 
## Splitrule:                        gini 
## OOB prediction error:             3.53 %

Predictions

## [1] 0.9509434

Confusion Matrix

##     predicted
## true    0    1
##    0 2541   32
##    1   99 1036

Final map

In the end, my model detected about 97 percent of the powerlines in this area.