Optimized Dilated Convolutional Neural Network with Quantum Self-Attention Based on Hyper Spectral Remote Sensing Image Classification

Hyperspectral remote sensing (RS) involves capturing and analyzing data across numerous, narrow spectral bands in the electromagnetic spectrum. With this technology, it is possible to identify what has been on the Earth surface with unprecedented depth that cannot be done with the mere multispectral images. A Hyperspectral sensor is able to capture hundreds of bands of data, and is therefore able to distinguish between very small differences in surface materials that we cannot even see or conventional multispectral sensors. Hyperspectral remote sensing is widely used in applications like agriculture, environmental monitoring, geology, and defence. To overcome the limitations of current approaches, new strategies are needed. In this manuscript Optimized Dilated Convolutional Neural Network with Quantum Self-Attention Based on Hyper Spectral Remote Sensing Image Classification (DCNN-QSA-PDOA) is proposed. The Indian Pine dataset is where the data is first gathered. The multidirectional shearlet transform based feature extraction (MDST) model is then utilized to extract the hyperspectral image (HSI) features. Following the completion of the feature extraction procedure, the Dilated Convolutional Neural Network with Quantum Self-Attention (DCNN-QSA) is trained on the feature extracted images and used to classify the category information. Here, (PDOA) is utilized for tuning hyper parameters of DCNN-QSA. The proposed DCNN-QSA-PDOA method provides 37.23%, 18.74% and 24.84% higher accuracy compared with existing method like Classifying images from hyper spectral RS using a DCNN-HIS, Dense residual three-dimensional convolutional neural networks are used for hyper spectral RS picture classification (DR-3DCNN-HIS), Fully convolutional neural networks are used for the categorization of hyper spectral remote sensing pictures (CNN-HIS) respectively.