Assessment of Accuracy for Soft Classification
Anil Kumar, Priyadarshi Upadhyay, Anil Kumar · 2020
Accuracy assessment of classified thematic maps generated using the classification algorithm is mandatory. For hard classification methods, there are well established accuracy assessment techniques. However, for soft classified outputs these established methods for accuracy assessment are not fully suitable and effective. Generally, soft classified outputs are converted to hard classified for accuracy assessment. In this chapter, some of the proposed methods for assessment of accuracy of soft classified outputs such as FERM and its different versions, entropy, RMSE, ROC, etc., have been explained. Further, it has been also covered how soft reference data can be used in FERM and its different versions.