Performance Analysis of YOLOv4 and SSD Mobilenet V2 for Foreign Object Debris (FOD) Detection at Airport Runway Using Custom Dataset

Abstract

Indonesia is a country that has heavy air traffic every day. Therefore, safety is a very important thing to pay attention to, one of them is the runway safety. The runway is an important component in aviation activities because aircraft use it for takeoff and landing. Foreign objects or FOD (Foreign Object Debris) could appear on the runway which can cause damage to the aircraft and may result in an accident. Therefore, we need a security system that can detect foreign objects in real-time. One approach that can be done is to use Computer Vision technology by using a camera. This method utilizes Artificial Intelligence (AI) technology for FOD detection. Various methods or algorithms have been developed for Computer Vision, SSD and YOLO are the most frequently used methods for real-time detection because of their high FPS and accuracy performance. Where in this study it was found that SSD MobileNet V2 can reach up to 12 FPS with mAP 0.5 value of 86.8% and for YOLOv4 can reach up to 31 FPS with mAP 0.5 value of 98.73%.