The Use of Band Selection in Weighted Linear Prediction for Hyperspectral Image Classification

The advancement of imaging procedures has made hyperspectral sensors fit for acquiring spectral data in many restricted and bordering bands, which brings about a high relationship between's neighboring bands and high information excess. It is important to decrease these bands previously advance analysis utilizing land cover classification and target location. Going for the classification undertaking, this paper displays another weighted technique for band selection, in view of band comparability estimation through Weighted Linear Prediction-based Band Selection (WLPBS). This method removes bands according to the correlation using a weighted linear prediction criterion. This makes it less demanding to accelerate the learning procedure and to enhance overall classification accuracy. Experimental results using Support a Vectors Machine (SVM) classifier on Airborne Visible Infrared Imaging Spectrometer (AVIRIS) datasets showed the effectiveness of our WLPBS method to improves the classification accuracy and to select informative and distinctive bands compared with the widely used Linear Prediction-based Band Selection (LPBS) technique and to those in the state-of-the-art. The maximum result obtained for classification of the selected band from Indian Pine and Salinas's datasets successively was 91.07% and 95.30% using WLPBS.