Object Detection and Classification from Hyper Spectral Images Using Adaptive Gaussian and Probabilistic Based K-Nearest Neighbour Classifier

Hyper Spectral Imagery (HSI) analysis plays a crucial role in remote sensing applications, such as object detection, identification and classification. Due to the complexity of HSI data, advanced analytical methods are required to produce accurate classification maps. A key research challenge in classification using HSI images is designing better algorithms with good efficiency capable of classifying the images captured by HSI techniques into various land cover types with precision. Processing the high dimensions of data, and class imbalance pose significant challenges. To address these issues, this research proposes a novel HSI classification model that integrates two methods – Principal Component Analysis for spectral based feature selection and a modified version of Gaussian-based method for enhancement of the features, and probabilistic model optimization for KNN classification. The main aim of the proposed method is to balance the tradeoff between accuracy and data loss in preprocessing by making use of both spatial and spectral features extraction and enhancement techniques.