The study of the effect of denoising on vectorized convolutional neural network

摘要: The remotely sensed high dimensional hyperspectral imagery is a single capture of scene at different spectral wavelengths. Since it contains an enormous amount information, has multiple areas application in the field remote sensing, forensic, biomedical etc. Hyperspectral images are very prone to noise due atmospheric effects and instrumental errors. In past, bands which were affected by discarded before further processing such as classification. Therefore along with relevant features present image lost. To avoid this, researchers developed many denoising techniques. goal technique remove effectively while preserving important features. Recently, convolutional neural network (CNN) servers bench mark on vision related task. Hence, can be classified using CNN. data fed pixel vectors thus called Vectorized Convolutional Neural Network (VCNN). this work determine effect VCNN. Here, VCNN functions classifier. For purpose comparison analyze trained raw (without denoising) denoised techniques as: Total Variation (TV), Wavelet, Least Square. performance classifier evaluated analyzing its precision, recall, F1-score. Also, based class-wise accuracies average for all methods been performed. From comparative classification result, observed that Square performs well