Neural Network

Multispectral image classification is a fundamental task in remote sensing, supporting applications such as land-cover mapping, agricultural monitoring, and environmental surveillance. Conventional classification methods, including Maximum Likelihood Classifier (MLC), Support Vector Machine (SVM), Decision Tree (DT), and Multi-Layer Perceptron (MLP), often face limitations when dealing with the high dimensionality and complex spectral characteristics of multispectral data. Recent advancements in deep learning have significantly improved the capability of remote sensing systems by enabling automatic extraction of high-level and discriminative features from raw data. In this study, we investigate the use of Deep Neural Networks (DNNs) for pixel-wise classification of multispectral satellite imagery. DNNs can effectively learn hierarchical feature representations, making them suitable for complex image analysis tasks. We propose a lightweight DNN architecture composed of six layers: an input layer representing spectral reflectance values across multiple bands, a fully connected layer, a batch normalization layer, a Rectified Linear Unit (ReLU) activation layer, a second fully connected layer, and a SoftMax output layer for multi-class classification. In the proposed approach, each pixel is represented as a vector of spectral reflectance values corresponding to different spectral bands.