PENGEMBANGAN SISTEM APLIKASI DETEKSI KELENGKAPAN PENAMBAT PADA JALAN REL BERBASIS YOU ONLY LOOK ONCE VERSI 8
DOI:
https://doi.org/10.19184/berkalafstpt.v2i1.891Keywords:
accuracy, detection, inspection, RMSE, You Only Look Once (YOLO), akurasi, deteksi, inspeksiAbstract
ABSTRAK
Gerakan dinamis roda kereta menimbulkan gaya lateral yang besar terhadap kedudukan rel. Selain itu aktivitas pencurian penambat oleh orang yang tidak bertanggung jawab juga dapat mempengaruhi kedudukan rel. Aplikasi “Rail Fasteners Detector” yang berbasiskan You Only Look Once versi ke-8 dapat mendeteksi kelengkapan dan membedakan jenis penambat kemudian direkam pada data logger. Hasil pengujian deteksi berdasarkan gambar menghasilkan nilai akurasi rata-rata tertinggi pada penambat E Clip yaitu 92,47% dengan nilai error (RMSE) juga paling rendah sebesar 0,77. Berdasarkan sumber video, hasil pengujian menghasilkan nilai akurasi tertinggi pada penambat Ka Clip yaitu sebesar 91,76% dengan nilai error (RMSE) juga paling rendah sebesar 0,1. Berdasarkan sumber webcam, hasil pengujian menghasilkan nilai akurasi tertinggi saat sudut pemasangan kamera sebesar 170° dengan kecepatan lori 15 km/jam dengan nilai akurasi 76,37%. Sementara nilai error (RMSE) paling rendah yaitu sudut pemasangan kamera sebesar 170° dengan kecepatan lori 25 km/jam memiliki nilai RMSE paling rendah sebesar 1,96.
ABSTRACT
The dynamic movement of the train wheels creates a large lateral force against the rail position. In addition, the theft of fasteners by irresponsible people can also affect the position of the rails. The "Rail Fasteners Detector" application based on the 8th version of You Only Look Once (YOLO) can detect completeness and distinguish the type of fasteners and then record them on the data logger. The results of the detection test based on the image produce the highest average accuracy value for the E Clip fastener, which is 92.47% with the lowest error value (RMSE) of 0.77. Based on the video source, the test results yielded the highest accuracy value for the Ka Clip fastener, which was 91.76% with the lowest error value (RMSE) of 0.1. Based on webcam sources, the test results yield the highest accuracy when the camera installation angle is 170° at a lorry speed of 15 km/hour with an accuracy value of 76.37%. While the lowest error value (RMSE), namely the camera installation angle of 170° at a lorry speed of 25 km/hour, has the lowest RMSE value of 1.96.
