Comprehensive Study of Applications of Supervised and Unsupervised Techniques on Spectral Data

Abstract

Recent advancements in computer vision, equipped with artificial intelligence-based models, have played a vital role in several areas, such as object detection, surveillance, autonomous vehicles, and augmented reality. These advancements have significantly improved the accuracy and efficiency of analyzing and interpreting data, expanding computer vision capabilities and applications across various sectors. Satellite image data comprises spatial and spectral information captured using different imaging techniques. Several research studies have been conducted, and challenges in the Precision Agriculture sector have been addressed using artificial intelligence (AI) based classification and segmentation approaches. Satellite data are available in multiple spectra, and they comprise information that can be extracted and can be used to address Precision Agriculture challenges. newlineThe agriculture sector faces numerous challenges, such as improper resource management, shrinking agricultural land, and diverse environmental conditions. These requirements are addressed using computer vision, AI, and integrated approaches. To address these issues, researchers have conducted numerous experiments with multiple data types comprising numeric, text, image, and video data. A significant focus was placed on obtaining advance crop yield estimates, which could be further used to monitor production and reduce agricultural losses, subsequently assisting in addressing broader challenges such as food security and sustainable agriculture. newlineIn our research, we have considered multispectral, hyperspectral, and thermal image data captured from different parts of the electromagnetic spectrum, each serving a unique purpose in observation and analysis. We have addressed challenges related to crop production, crop classification, and intrusion detection. newline There is an intrinsic requirement for extensive labeled data to train AI algorithms. The algorithm performance is substantially impacted without a significant quantity of labeled data. Therefore,