Perancangan Deteksi Rintangan Untuk Traktor Menggunakan Kamera Stereo Berbasis Estimasi Kedalaman Dan Filter Kalman

Purwantoro, Ghani Indrastoto (2025) Perancangan Deteksi Rintangan Untuk Traktor Menggunakan Kamera Stereo Berbasis Estimasi Kedalaman Dan Filter Kalman. Other thesis, Institut Teknologi Sepuluh Nopember.

Text 5009211015-Undergraduate_Thesis.pdf - Accepted Version Restricted to Repository staff only Download (4MB) | Request a copy

Abstract

Sebagai respons terhadap kebutuhan peningkatan produktivitas dan otomatisasi di sektor pertanian, penelitian ini merancang sistem deteksi rintangan untuk traktor menggunakan kamera stereo, estimasi kedalaman, dan filter kalman. Sistem ini bertujuan untuk mengukur jarak antara traktor yang bergerak dengan objek statis di depannya secara akurat melalui algoritma berbasis stereo vision. Metode yang digunakan meliputi pra-pemrosesan citra, segmentasi area objek, ekstraksi informasi disparitas, dan perhitungan jarak objek menggunakan geometri stereo. Untuk meningkatkan akurasi dan mengurangi gangguan noise pengukuran, filter Kalman diterapkan pada data kedalaman. Hasil training dataset menggunakan model YOLOv5-nano sebagai pendeteksi objek mendapatkan nilai mAP sebesar 82% dan pengujian secara real plant dan didapatkan nilai presisi sebesar 83%. Sistem estimasi kedalaman menunjukkan average error pengukuran di bawah 5% untuk jarak hingga 7,5 meter. Hasil implementasi menunjukkan bahwa kombinasi kamera stereo dan filter kalman efektif dalam meningkatkan keandalan sistem deteksi rintangan sebagai fondasi bagi traktor otonom di masa depan.
==================================================================================================================================
In response to the need for increased productivity and automation in the agricultural sector, this study designs an obstacle detection system for tractors using stereo cameras, depth estimation, and a Kalman filter. The system aims to accurately measure the distance between a moving tractor and static objects ahead through a stereo vision-based algorithm. The methods employed include image pre-processing, object area segmentation, disparity information extraction, and object distance calculation using stereo geometry. To improve accuracy and reduce noise measurement, a Kalman filter is applied to the depth data. Training results using the YOLOv5-nano model as an object detector achieved a mean Average Precision (mAP) of 82%, and real-plant testing yielded a precision score of 83%. The depth estimation system demonstrated an average measurement error of less than 5% for distances up to 7.5 meters. The implementation results show that the combination of stereo cameras and the Kalman filter effectively enhances the reliability of the obstacle detection system as a foundation for future autonomous tractors.

Item Type:	Thesis (Other)
Uncontrolled Keywords:	deteksi rintangan, estimasi kedalaman, kalman filter, kamera stereo, traktor, YOLOv5. ; depth estimation, kalman filter, obstacle detection, stereo camera, tractor
Subjects:	T Technology > T Technology (General) T Technology > T Technology (General) > T59.7 Human-machine systems. T Technology > TL Motor vehicles. Aeronautics. Astronautics > TL152.8 Vehicles, Remotely piloted. Autonomous vehicles.
Divisions:	Faculty of Industrial Technology and Systems Engineering (INDSYS) > Physics Engineering > 30201-(S1) Undergraduate Thesis
Depositing User:	Ghani Indrastoto Purwantoro
Date Deposited:	05 Aug 2025 03:43
Last Modified:	11 Aug 2025 02:57
URI:	http://repository.its.ac.id/id/eprint/126396

Actions (login required)

View Item