An Intelligent deep learning based hyperspectral Signal classification scheme for complex measurement systems

Advanced Hyperspectral Imaging (HIS) systems generate massive volumes of datasets that can provide significant details, when appropriately mined. However, analysis and the interpretation of such huge volume of data is a challenging task to accomplish. Therefore, Deep Learning (DL) methods are highly helpful in solving conventional image processing tasks and it also offers new stimulating issues in spatial-spectral domain. Since effective ground feature extraction from HSI is a challenging research domain, the current research article designs an Intelligent DL-based Hyperspectral Signal Classification (IDL-HSSC) scheme for complex measurement systems. The aim of the proposed IDL-HSSC technique is to classify the HSI under appropriate class labels to understand the ground features. Besides, IDL-HSSC technique involves the design of Tree Growth Algorithm (TGA) with SqueezeNet model for the extraction of feature vectors, where TGA is employed to select the hyperparameters. Moreover, Biogeography-Based Optimization (BBO) with Cascaded Forward Neural Network (CFNN) is also employed as a classifier to categorize the images under appropriate class labels. Both TGA and BBO algorithms are designed for the optimization of parameters used in SqueezeNet and CFNN techniques which in turn helps in accomplishing the maximum classification outcomes. In order to ensure the proficient performance of the proposed IDL-HSSC technique, a wide range of experiments was conducted on diverse benchmark datasets. The experimental outcomes established the supreme performance of the proposed IDL-HSSC technique over recent state-of-the-art methods.