Surveying Multi-object Classification Methods for UAV Imagery Analysis
AFRICAN JOURNAL OF APPLIED RESEARCH · 18.11.2026 · 12 (8) · 147-175
DOI: 10.26437/ngfzsa84 ↗Kurz gefasst
Fachartikel in der Zeitschrift AFRICAN JOURNAL OF APPLIED RESEARCH, online erschienen am 18.11.2026. Verfasst von S. V. Kuntawar und 1 weiteren Autor:innen. Die Arbeit nennt 65 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY-NC 4.0).
Abstract (Originalsprache)
Purpose
This paper presents a comprehensive review of multi-object classification methods for Unmanned Aerial Vehicle (UAV) imagery by examining the evolution of traditional machine learning and modern deep learning techniques. The objectives are to determine existing approaches, compare their performance, identify research gaps, and develop a guide for accurate and efficient UAV image analysis. Design/Methodology/Approach: The research offered a comprehensive, cutting-edge object categorisation system for aerial images that incorporates single-stage, two-stage, and advanced-stage detectors. Many recent state-of-the-art object classification algorithms are designed to classify one object at a time. This necessitates new processing and analysis techniques that can fully utilise the enormous volume of gathered information. Research Limitation: The review is limited to published literature and benchmark datasets available during the study period. Variations in datasets, evaluation protocols, experimental settings, and reported performance metrics limit direct comparison across methods.
Findings
The review found that deep learning-based methods significantly outperform traditional machine learning techniques in detection accuracy, robustness, and feature-learning capabilities. CNN-based architectures dominate current UAV object classification research, while attention mechanisms and multimodal approaches further improve performance in challenging environments. The analysis also identified persistent challenges, including object occlusion, scale variation, complex backgrounds, limited annotated datasets, computational complexity, and inconsistent reporting of evaluation metrics. Practical Implication: The findings provide valuable guidance for researchers, practitioners, and developers in selecting suitable object classification techniques for UAV applications. Social Implication: It has an impact on precision agriculture, environmental monitoring, disaster management, traffic surveillance, urban planning, and security Originality/Value: Unlike existing surveys, it covers everything from basic traditional algorithms to advanced deep learning algorithms, methodologies, frameworks, new hypotheses, procedures, and applications, although earlier studies focused on single-object detection, particularly a specific target.
Zitieren
APA
Kuntawar, S. V., & Khurshid, A. A. (2026). Surveying Multi-object Classification Methods for UAV Imagery Analysis. AFRICAN JOURNAL OF APPLIED RESEARCH, 12(8), 147-175. https://doi.org/10.26437/ngfzsa84
Vancouver
Kuntawar SV, Khurshid AA. Surveying Multi-object Classification Methods for UAV Imagery Analysis. AFRICAN JOURNAL OF APPLIED RESEARCH. 2026;12(8):147-175. doi:10.26437/ngfzsa84
BibTeX
@article{Kuntawar2026,
author = {Kuntawar, S. V. and Khurshid, A. A.},
title = {Surveying Multi-object Classification Methods for UAV Imagery Analysis},
journal = {AFRICAN JOURNAL OF APPLIED RESEARCH},
year = {2026},
volume = {12},
number = {8},
pages = {147--175},
doi = {10.26437/ngfzsa84}
}
EUVIMED