Performance évaluation of spectral angle mapper and spectral correlation mapper classifiers over multiple remote sensor data

The concept of remote sensing is to study the properties of Earth's objects by recording its reflection. Development in remote sensing has been consistent in the past few decades ranging from optical to microwave domains. Classification enables us to extract information of the objects existing on the image surface. Selection of suitable classification algorithm is dependent on the environmental conditions and the distribution of land units. An attempt has been made to study the performance of Spectral Angle Mapper (SAM) and Spectral Correlation Mapper (SCM) classifiers over multiple data sets. Also, the study tries to emphasize the performance of the SAM and SCM as the type and size of the data set are varied. Accuracy assessment was performed to study the potential of each classifier. The results of the classifiers have been analyzed with respect to the class separabilities and spectral characteristics of the data sets. This study analyzes the effect of spectral overlapping on individual class classification accuracy. SCM classifier extracted information more effectively in situations of severe class overlapping conditions.