International Research Journal of Engineering and Technology (IRJET)
e-ISSN: 2395-0056
Volume: 13 Issue: 06 | Jun 2026
p-ISSN: 2395-0072
www.irjet.net
Integrating AI and Computer Vision for Effective Foreign Object Detection on Airport Runways Dr. CH. Sarada Devi, M. Tech, Ph. D1, Ms. D. Rejitha, M. E, (Ph. D)2, Abinaya. S3 1Head of the Department, Department of Computer Science and Engineering, Meenakshi College of Engineering
K. K. Nagar Chennai, India
2Assistant Professor Department of Computer Science and Engineering, Meenakshi College of Engineering
K. K. Nagar Chennai, India
3Department of Computer Science and Engineering, Meenakshi College of Engineering K. K. Nagar Chennai, India
--------------------------------------------------------------------------------***-----------------------------------------------------------------------models. Abstract-The FOD-R Dataset consists of images To facilitate research in this domain, the FOD-A representing common types of Foreign Object Debris dataset was introduced as a comprehensive (FOD) typically found on airport runways and taxiways. benchmark for training and evaluating machine The dataset is primarily annotated with bounding boxes learning and computer vision algorithms. The object to support object detection tasks. An enhanced version, categories included in the dataset were selected known as the FOD-xR Dataset, incorporates an based on recommendations from the Federal additional feature in the form of color-coded labels that Aviation Administration (FAA) and findings from indicate the difficulty of debris removal. Green labels previous research, ensuring that the dataset represent lightweight objects that are easier to remove, represents a broad range of debris commonly found whereas red labels indicate heavier objects that require on airport runways and taxiways. Images were greater effort. This centralized labeling approach collected under varying environmental conditions, improves coordination among maintenance teams by including different lighting levels and rainfall enabling efficient planning and management of FOD scenarios, to simulate real-world airport removal operations. As a result, organizations can environments and improve the generalization reduce operational delays, minimize aircraft damage, capability of detection models. and improve overall airport safety. The FOD-R and FODxR datasets serve as valuable resources for developing The FOD-A dataset was also designed with intelligent FOD detection and removal systems. scalability in mind, allowing new images and object categories to be incorporated while preserving Index Terms-Foreign Object Debris (FOD), annotation consistency. Multiple versions of the centralized management, dataset, bounding box dataset are maintained to ensure that it remains annotation, object detection, color coding. comprehensive and up to date. Researchers can evaluate detection algorithms using a standardized I. INTRODUCTION version of the dataset and subsequently utilize newer releases to develop enhanced models capable Foreign Object Debris (FOD) poses a significant threat of recognizing additional FOD categories. Figure 1 to aviation safety, as it can lead to aircraft damage, illustrates a sample image from the FOD-A dataset operational disruptions, injuries, and even fatalities. with annotated bounding boxes. The financial losses associated with FOD incidents are also substantial due to aircraft repairs, flight delays, Although several publicly available image datasets and maintenance costs. To address these challenges, contain common object categories such as vehicles, researchers have increasingly explored the use of furniture, and household items, they do not Machine Learning (ML) and Computer Vision (CV) adequately represent the specialized debris techniques for the automatic detection and encountered in airport environments. Important classification of FOD. The development of accurate and FOD categories, including aircraft components, reliable ML-based detection systems requires highmaintenance tools, luggage items, and metallic quality datasets containing diverse FOD images fragments, are largely absent from these generalcaptured under realistic operating conditions. Such purpose datasets. Therefore, only limited datasets play a vital role in improving the accuracy, comparisons with conventional object detection robustness, and efficiency of intelligent FOD detection datasets are presented, as they cannot effectively
© 2026, IRJET
|
Impact Factor value: 8.315
|
ISO 9001:2008 Certified Journal
|
Page 922