Inherent Feature Extraction and Soft Margin Decision Boundary Optimization Technique for Hyperspectral Crop Classification

Abstract

Crop productivity and disaster management can be enhanced by employing hyperspectral images. Hyperspectral imaging is widely utilized in identifying and classifying objects on the ground surface for various agriculture application uses such as crop mapping, flood management, identifying crops damaged due to flood/drought, etc. Hyperspectral imaging-based crop classification is a very challenging task because of spectral dimensions and poor spatial feature representation. Designing efficient feature extraction and dimension reduction techniques can address high dimensionality problems. Nonetheless, achieving good classification accuracies with minimal computation overhead is a challenging task in Hyperspectral imaging-based crop classification. In meeting research challenges, this work presents Hyperspectral imaging-based crop classification using soft-margin decision boundary optimization (SMDBO) based Support Vector Machine (SVM) along with Image Fusion-Recursive Filter (IFRF) and Inherent Feature Extraction (IFE). In this work, IFRF is used for reducing spectral features with meaningful representation. Then, IFE is used for differentiating physical properties and shading elements of different objects spatially. Both spectral and spatial features extracted are trained using SMDBO-SVM for performing hyperspectral object classification. Using SMDBO-SVM for Hyperspectral object classification aid in addressing class imbalance issues; thus, the proposed IFE-SMDBO-SVM model achieves better accuracies and with minimal misclassification in comparison with state-of-art statistical and Deep Learning (DL) based Hyperspectral object classification model using standard dataset Indian Pines and Pavia University.

Hyperspectral object classification using hybrid spectral-spatial fusion and noise tolerant soft-margin technique

“…where 𝐽 ̂𝑙 defines the 𝑙 th subcluster, 𝐴 defines the band size of each cluster, ⌈ . Then, for obtaining shading and reflectance features of each object optimization is done using an intrinsic feature extraction mechanism at each subcluster 𝐽 ̂𝑙 as defined in (5).…”

“…A preprocessing technique called image-fusion enhances an image's spectral and spatial resolution. As illustrated in Figure 1, the image fusion method finds application in many domains, including medical image visualization, machine vision [5], bioinformatics security, land classification, navigation, variation identification, digital imaging, military applications, satellite, and aerial imaging [6], [7], robotic vision, food microbe detection [8], photography, and surveillance [9]. The goal of this study is to examine current advancements in image-fusion techniques used in HSI-based object classification methods, pinpoint issues and difficulties encountered, and provide a successful feature-fusion method that preserves intrinsic object characteristics in both spatial and spectral domains.…”

Application of deep learning methods for automated analysis of retinal structures in ophthalmology

“…In modern ophthalmology, advanced deep learning technologies are being actively implemented [1], [2] for automated analysis of retinal structures. Some of the key methods that have attracted the attention of researchers and clinicians are EfficientNet [3], [4] and DenseNet [5], [6]. These methods provide powerful tools for image processing [7], [8] with a high degree of efficiency.…”

A hybrid spectral-spatial fusion technique for hyperspectral object classification

“…Image-fusion is a preprocessing method that improves both the spectral as well as the spatial resolution of an image. Image fusion method are utilized in a variety of fields as shown in Figure 1, including such as the visualization in the medical images, machine-vision [5], security in bioinformatics, classification of land, navigation, variation identification, digital imaging, military applications, satellite, and aerial imaging [6], [7], robotic vision, detection of food microbes [8], photography and surveillance [9]. The purpose of this research is to investigate recent developments in the field of image-fusion technique applied for object classification approaches using HSI and identify the problems and challenges encountered and present an effective feature fusion technique that retain object intrinsic characteristic both spatially and as well as spectrally.…”