Airport Runway Foreign Object Debris (FOD) Detection Based on YOLOX Architecture

Abstract

Today, every airport manager in various countries has tightened runway security to avoid the entry of foreign objects that can endanger passengers and aircraft both when landing and taking off. Inspection and supervision of the runway must be carried out regularly. However, there are still many airports that carry out inspections and supervision by human labor without any technological support. Whereas inspection and supervision using human labor takes a relatively long time and is prone to errors, especially in bad weather and limited visibility. Technological developments in runway security using radar are one of the solutions. However, radar technology is relatively expensive, so many airport managers use computer vision because it is considered cheaper and more accurate. The use of computer vision has grown rapidly in monitoring FOD on aircraft runways. Our method is an impovement of the YOLOX architecture by moving output objects to branch classes. Our method got a MAP score of 0.832 which has an increase in score of 0.021 from the previous method in detecting FOD in classes of people, vehicles, birds, cats and dogs.