3 Results
3.1 Vegetation
Classification
3.1.1 Classification Accuracy
Assessment and
Comparisons
All methods achieved an overall
accuracy of at least 79% and Kappa coefficient above 0.74, indicating
high repeatability and accuracy (Figure 5). Among three methods,
SVM demonstrated superior
performance, exhibiting the highest overall accuracy of 84.0% and Kappa
coefficient of at least 0.81. This represents a 5% and 10% higher
accuracy compared to the other algorithms, respectively. The confusion
matrix of SVM classification can be found in Table 1. It reveals that
Caragana, Poplar, and Grass have a high classification accuracy of more
than 89%. The classification accuracy of Salix and Artemisia is
relatively low, at 71.43% and 72.92% respectively; while Corn is the
lowest, at 68.54%.