Hybrid classification using combination of optimized spectral angle mapping algorithm and interpolation method on multispectral and hyper spectral image

A growing number of studies in recent years has focused on improvement of performance of classification algorithm on hyper-spectral image; this provides a scientific basis for efficient object detection. This paper tries to improve the performance of spectral angle mapping algorithm to classify the hyper-spectral image. The proposed method uses combination of spectral angle mapping algorithm and interpolation method with supervised clustering techniques for efficient object detection and region finding. Spectral angle mapping algorithm is used for finding pure pixel, thereby reducing the probability of false object detection due to geometric errors. The experimental results show that, proposed hybrid technique reduces the probability of false object detection with inbuilt radiometric error enhancement capability for hyper-spectral image.