Pengembangan Point of Interest Pada Indoor Drone Menggunakan Feature Matching

Swista, Maharani Wisudawati (2025) Pengembangan Point of Interest Pada Indoor Drone Menggunakan Feature Matching. Masters thesis, Institut Teknologi Sepuluh Nopember.

Text 6022222013-Master _Thesis.pdf Download (5MB)

Abstract

Penelitian ini membahas sistem navigasi drone dalam ruangan menggunakan teknologi Point of Interest (POI) berbasis penginderaan visual. Drone jenis MAV (Micro Aerial Vehicle) digunakan untuk mengorbit objek target secara otonom tanpa bergantung pada GPS, yang sering terganggu dalam lingkungan tertutup. Algoritma ORB dipilih untuk mendeteksi dan mencocokkan citra objek 3D secara waktu nyata karena ketahanannya terhadap noise. Target pengujian berupa lingkaran berdiameter 10 cm, 20 cm, dan 30 cm, dengan variasi menggunakan kontur dan tanpa kontur. Hasil menunjukkan bahwa penggunaan kontur membantu sistem visual mempertahankan struktur distribusi titik estimasi, meskipun akurasi tetap rendah. Untuk diameter 10 cm, rata-rata error mencapai 0,24 meter dengan RMSE 0,28 meter. Sementara itu, pengukuran tanpa kontur
menghasilkan distribusi titik yang lebih acak dan sulit dikenali. Kamera Intel RealSense T265 yang digunakan untuk pemetaan posisi menunjukkan adanya penyimpangan (drift), terutama dalam kondisi tanpa kontur. Pendeteksian menggunakan ORB cukup efektif, ditunjukkan dengan akurasi tertinggi pada frame ke-16 (error 7,07%). Namun, frame lain seperti ke-39 dan ke-59 menunjukkan error besar hingga 102,94%. Kesimpulannya, sistem navigasi visual ini masih membutuhkan peningkatan untuk aplikasi presisi di lingkungan dalam ruangan.
======================================================================================================================================
This study explores an indoor drone navigation system utilizing Point of Interest (POI) technology based on visual sensing. A Micro Aerial Vehicle (MAV) is employed to autonomously orbit a target object without relying on GPS, which is often disrupted in enclosed environments. The ORB algorithm is selected for detecting and matching 3D object images in real time due to its robustness against noise. The test targets are circular objects with diameters of 10 cm, 20 cm, and 30 cm, with variations using contour and no contour. The results show that the use of contours helps the visual system maintain a more structured distribution of estimated position points, although the overall accuracy remains low. For the 10 cm diameter target, the average error reached 0.24 meters with an RMSE of 0.28 meters. In contrast, measurements without contours resulted in more scattered and unrecognizable point distributions. The Intel RealSense T265 camera used for position mapping exhibited significant drift, especially in the absence of contours. ORB-based detection proved to be relatively effective, with the highest accuracy observed at frame 16 (7.07% error). However, other frames such as frame 39 and frame 59 showed much higher errors, up to 102.94%. In conclusion, while the visual-based navigation system shows potential, further optimization is required for precise indoor applications.

Item Type:	Thesis (Masters)
Uncontrolled Keywords:	Point of Interest, pencocokan fitur, objek, drone, T265 Realsense Point of Interest, feature matching ORB, drone, object, T265 Realsense
Subjects:	T Technology > T Technology (General) T Technology > TK Electrical engineering. Electronics Nuclear engineering > TK3070 Automatic control T Technology > TK Electrical engineering. Electronics Nuclear engineering > TK6592.A9 Automatic tracking.
Divisions:	Faculty of Intelligent Electrical and Informatics Technology (ELECTICS) > Electrical Engineering > 20101-(S2) Master Thesis
Depositing User:	Maharani Wisudawati Swista
Date Deposited:	29 Jul 2025 02:51
Last Modified:	29 Jul 2025 02:51
URI:	http://repository.its.ac.id/id/eprint/122453

Actions (login required)

View Item