FHIC: Fast Hyperspectral Image Classification Model Using ETR Dimensionality Reduction and ELU Activation Function

Hyperspectral images (HSIs) are typically utilized in a wide variety of practical applications. HSI is replete with spatial and spectral information, which provides precise data for material detection. HSIs are characterized by a high degree of variations and undesirable pixel distributions, providing major processing challenges. This article introduces the fast hyperspectral image classification (FHIC) model, a rapid model for classifying HSIs and resolving their associated challenges. It uses the enhancing transformation reduction (ETR) method to address the HSI difficulties and enhance classes’ differentiation. It also uses exponential linear units (ELUs) to smooth and speed the classification processing. The structure of the FHIC model is designed to be very flexible and suitable for a range of HSIs. The model reduced execution time and RAM consumption, and provided superior performance compared to seven of the most advanced analysis models for three well-known HSIs. In some cases, it was 60% faster than other models. In addition, this work presents a new and highly effective method for measuring the performance of the compared models in terms of their accuracy and processing speed to provide an easy evaluation method. The code of the FHIC model is available at this link: https://github.com/DalalAL-Alimi/FHIC.

I. Introduction

Hyperspectral imaging (HSI) uses specialized sensors to simultaneously collect data at various narrow wavelengths. The gathered data is organized into a “hyperspectral cube,” which has three dimensions: two of which indicate the scene’s spatial extent and the third its spectral content. The rapid advancement of remote sensing technology has greatly increased the spatial resolution of HSI datasets, significantly enhancing the ability of HSI datasets to express unique objects accurately. This advancement has spread to the industrial, scientific, and military fields. The acquired images have many challenges and problems, making researchers face difficulties in extracting the HSI features and targets. These features and challenges of the HSI field make it attractive for many researchers. The enormous dimensionality of HSI brought redundant information and increased consumption of computing and resources. In addition, besides the redundant information, the HSI has challenges and many mixed pixels. Mixed pixels frequently correspond to multiple categories and cause significant challenges for classification. Due to a flaw or issue with the detectors, dead pixels represent zero or missing values [1], [2], [3]. There are small ratios between the numbers of samples in many classes due to manually labeling HSI samples [4], [5], [6]. Moreover, due to atmospheric variability, HSI includes undesirable data such as outliers and noise [7], [8].