Convolutional Neural Network Implementation in Detection of Approach Lights Lighting Condition

Abstract

Approach Light is an aircraft visual landing aid in a certain form of lighting to assist pilots when landing an aircraft in the dark or bad weather (below average visibility) in order to land safely. With the important role of the Approach Light in the aircraft landing process, the ON and OFF lighting condition of the Approach Light is necessary to be monitored. The design of this research uses artificial intelligence technology that can determine whether the lights on the Approach Light are in ON or OFF condition using camera's image capture. To find out whether the lights are on or not, Convolutional Neural Network is implemented in this monitoring technique to process image classification oh the lights. It can also send evidence in the form of captured images classified on the website as evidence of monitoring results that can be confirmed by technicians if any inappropriate classification results occurred. The results showed that the classification results for each brightness step obtained average values of 95% in accuracy, 90% in prediction precision, and 98% in prediction sensitivity. According to this good result of values, it is expected to give positive contribution for the technicians so that flight operations disruption can be minimized.