OBJECT DETECTION IN MULTISPECTRAL AERIAL IMAGES USING MACHINE LEARNING ALGORITHMS: A COMPARATIVE STUDY OF K-MEANS AND FUZZY C-MEANS APPROACHES

Abstract

Object detection within multispectral aerial images holds immense importance across a spectrum of applications, encompassing agriculture, environmental monitoring, surveillance, and urban planning. This research presents a comprehensive inquiry into the utilization of machine learning algorithms for detecting objects within multispectral aerial images. The approach commences by exploring a gamut of preprocessing methods, including histogram equalization, erosion, dilation, opening, closing, grayscale transformation, contrast enhancement, sharpening, and denoising. These preprocessing techniques play a pivotal role in augmenting the quality of multispectral images, thereby amplifying the efficacy of ensuing object detection algorithms. Following this, the study delves into the execution and assessment of the K-Means and Fuzzy C- Means algorithms for object detection. To gauge the performance of the proposed methodologies, a stringent evaluation involving accuracy, precision, recall, and F1-score metrics is employed. The empirical findings divulge the implications of preprocessing techniques and the subsequent algorithmic choices on the outcomes of detection. By contrasting the outcomes of the K-Means and Fuzzy C-Means methodologies, an analysis is conducted to elucidate their respective competencies and limitations within object detection contexts. This research accentuates the pivotal role of preprocessing and algorithm selection in achieving precise object detection within multispectral aerial images. By elucidating the strengths and constraints of the K-Means and Fuzzy C-Means techniques, this study lays the groundwork for future advancements in multispectral image analysis through the prism of machine learning algorithms.