PENERAPAN MACHINE LEARNING UNTUK MEMANTAU KERUSAKAN KONTAINER PADA GATE TERMINAL PETIKEMAS MAKASSAR NEW PORT

Abstract

This research was conducted to design and develop a system capable of automatically monitoring and determining roof damage on containers before they enter the container yard. The monitoring is carried out at the gate area by applying a Machine Learning method, namely You Only Look Once (YOLO). Three types of damage are being monitored: rust, holes, and dents. Human error, extreme weather conditions, or collisions with other objects usually cause this damage. The results of the study show that YOLO in monitoring and identifying container damage achieved an 88.8% detection accuracy in the morning with a detection process duration of 31.5 seconds, a 90.9% detection accuracy during the day with a detection process duration of 31.1 seconds, an 87.7% detection accuracy in the afternoon with a detection process duration of 32.6 seconds, and an 83.6% detection accuracy at night with a detection process duration of 33.2 seconds. The low detection accuracy at night, which is only 83.6% is caused by less than optimal lighting, in contrast to daytime lighting. The real- time monitoring system functions quite well, as it can provide documentation evidence and continuously notify the gate inspector regarding the physical condition of the container roof. Keywords: Container, Gate, Machine Learning, Real-Time Monitoring System, You Only Look Once