Pendekatan Algoritma You Only Look Once (YOLO) untuk Mendeteksi Cavity dan Utilitas Bawah Tanah pada Data Ground Penetrating Radar

Mulia, Aldila Putri (2026) Pendekatan Algoritma You Only Look Once (YOLO) untuk Mendeteksi Cavity dan Utilitas Bawah Tanah pada Data Ground Penetrating Radar. Other thesis, Institut Teknologi Sepuluh Nopember.

Text 5017211038-Undergraduate_Thesis.pdf - Accepted Version Restricted to Repository staff only Download (3MB) | Request a copy

Abstract

Cavity dan utility merupakan komponen penting dalam infrastruktur perkotaan yang perlu dideteksi secara dini untuk mencegah potensi kegagalan struktur. Metode Ground Penetrating Radar (GPR) merupakan pendekatan non-destruktif yang efektif untuk investigasi bawah permukaan, namun interpretasi radargram masih bergantung pada subjektivitas analis. Penelitian ini bertujuan mengevaluasi kinerja algoritma You Only Look Once (YOLO) dalam mendeteksi cavity dan utility dari citra GPR serta menilai akurasi dan konsistensi interpretasi objek bawah permukaan menggunakan model hasil pelatihan. Hasil evaluasi menunjukkan bahwa model YOLOv8n dan YOLOv8s mampu mendeteksi objek bawah permukaan pada data GPR, dengan YOLOv8n menunjukkan kinerja yang lebih unggul berdasarkan nilai performance metrics, yaitu Precision 0,788, Recall 0,805, mAP50 0,831, dan F1 Score 0,796, yang mencerminkan keseimbangan yang baik antara ketepatan deteksi dan sensitivitas. Pengujian data lapangan menunjukkan bahwa tingkat ketepatan dan konsistensi deteksi dipengaruhi oleh kualitas citra GPR, khususnya kejelasan refleksi dan tingkat noise; pada kondisi variasi pola refleksi yang lebih kompleks, model YOLOv8s menunjukkan performa deteksi yang lebih konsisten dibandingkan YOLOv8n. Secara keseluruhan, integrasi data GPR dan algoritma YOLO berpotensi meningkatkan objektivitas interpretasi radargram serta menyediakan pendekatan komputasi yang andal untuk deteksi objek bawah permukaan secara otomatis apabila didukung oleh kualitas data yang memadai.
=====================================================================================================================================
Subsurface cavities and underground utilities are essential components of urban infrastructure that require early detection to prevent potential structural failures. The Ground Penetrating Radar (GPR) method provides an effective non-destructive approach for subsurface investigation; however, radargram interpretation remains largely dependent on analyst subjectivity. This study aims to evaluate the performance of the You Only Look Once (YOLO) algorithm in detecting cavities and underground utilities from GPR images and to assess the accuracy and consistency of subsurface object interpretation using trained models. The evaluation results show that both YOLOv8n and YOLOv8s are capable of detecting subsurface objects in GPR data, with YOLOv8n achieving higher values in performance metrics, including a Precision of 0.788, Recall of 0.805, mAP50 of 0.831, and an F1 Score of 0.796, indicating a strong balance between detection accuracy and sensitivity. Field data testing reveals that detection accuracy and consistency are influenced by GPR image quality, such as reflection clarity and noise levels. Under conditions with more complex reflection pattern variations, the YOLOv8s model demonstrates more consistent detection performance than YOLOv8n, despite both models being able to interpret subsurface objects. Overall, the integration of GPR data and the YOLO algorithm shows strong potential for enhancing the objectivity of radargram interpretation and provides a reliable computational approach for automated subsurface object detection when supported by adequate data quality.

Item Type:	Thesis (Other)
Uncontrolled Keywords:	Cavity, Deteksi Objek, Ground Penetrating Radar, Utility, YOLOv8n, YOLOv8s. Cavity, Ground Penetrating Radar (GPR), Object Detection, Subsurface Utilities, YOLOv8n, YOLOv8s.
Subjects:	Q Science Q Science > QE Geology > QE33.2.R33 Ground penetrating radar
Divisions:	Faculty of Civil, Environmental, and Geo Engineering > Geophysics Engineering > 33201-(S1) Undergraduate Theses
Depositing User:	Aldila Putri Mulia
Date Deposited:	03 Feb 2026 02:23
Last Modified:	03 Feb 2026 02:23
URI:	http://repository.its.ac.id/id/eprint/131809

Actions (login required)

View Item