Rancang Bangun Aplikasi Mobile dan Model Aproksimasi Massa Tandan Buah Segar Pada Sistem Taksasi Produksi Kelapa Sawit

Baihaqi, Faiq Lidan (2026) Rancang Bangun Aplikasi Mobile dan Model Aproksimasi Massa Tandan Buah Segar Pada Sistem Taksasi Produksi Kelapa Sawit. Other thesis, Institut Teknologi Sepuluh Nopember.

Text 5025221294-Undergraduate_Thesis.pdf Restricted to Repository staff only Download (8MB) | Request a copy

Abstract

Industri kelapa sawit Indonesia menghadapi tantangan pada proses taksasi produksi yang masih bergantung pada metode konvensional yang rentan terhadap subjektivitas dan tidak efisien di lapangan. Penelitian ini mengembangkan dua kontribusi dalam sistem taksasi produksi kelapa sawit: aplikasi mobile Android dan model aproksimasi massa Tandan Buah Segar (TBS). Aplikasi dibangun menggunakan Kotlin dengan pola clean architecture dan mekanisme offline-first berbasis WorkManager, mencakup visualisasi peta sebaran pohon interaktif, rekonstruksi objek 3D, dan akuisisi data visual TBS. Model aproksimasi massa TBS dikembangkan menggunakan deteksi objek YOLOv8s-OBB dengan formula prolate spheroid, diintegrasikan ke dalam aplikasi dalam format TensorFlow Lite dengan memanfaatkan GPU delegate berpresisi float16. Dataset terdiri dari 855 gambar dengan tiga kelas dari 10 pohon kelapa sawit varietas Tenera pada dua jarak perekaman. Seluruh 31 skenario uji fungsional berhasil, dengan waktu rendering peta dan visualisasi 3D berada di bawah threshold 30 detik dan hanya 4,5% frame janky dari total 1.132 frame yang di-render. Model YOLOv8s-OBB mencapai mAP@0.5:0.95 sebesar 0,7942 secara keseluruhan, dengan 0,7883 pada kelas matang dan 0,8875 pada kelas penggaris. Pengujian estimasi massa terhadap 20 video evaluasi menghasilkan MAPE sebesar 44,88% dan MAE sebesar 3,96 kg. Pengujian implementasi pada aplikasi menunjukkan waktu pemrosesan mendekati waktu nyata berkat GPU delegate, meskipun masih terdapat selisih terhadap hasil notebook akibat perbedaan laju ekstraksi frame antara kedua implementasi.
====================================================================================================================================
Indonesia’s palm oil industry faces challenges in the production estimation process, which still relies on conventional methods that are prone to subjectivity and inefficient in the field. This research makes two contributions to the oil palm production estimation system: an Android mobile app and a Fresh Fruit Bunch (FFB) mass approximation model. The app was built using Kotlin with a clean architecture pattern and an offline-first mechanism based on WorkManager, featuring interactive tree distribution map visualization, 3D object reconstruction, and FFB visual data acquisition. The FFB mass approximation model was developed using YOLOv8s-OBB object detection with a prolate spheroid formula, integrated into the app in TensorFlow Lite format by leveraging a float16-precision GPU delegate. The dataset consists of 855 images with three classes from 10 Tenera-variety oil palm trees at two recording distances. All 31 functional test scenarios were successful, with map rendering and 3D visualization times remaining below the 30-second threshold and only 4.5% of the total 1,132 rendered frames exhibiting jankiness. The YOLOv8s-OBB model achieved an overall mAP@0.5:0.95 of 0.7942, with 0.7883 for the ripe class and 0.8875 for the ruler class. Mass estimation testing on 20 evaluation videos yielded a MAPE of 44.88% and an MAE of 3.96 kg. Testing of the implementation in an application showed processing times approaching real-time thanks to GPU delegation, although there was still a discrepancy with the notebook results due to differences in frame extraction rates between the two implementations.

Item Type:	Thesis (Other)
Uncontrolled Keywords:	Aplikasi mobile Android, aproksimasi massa TBS, Clean architecture, Offline-first, Prolate spheroid, Taksasi produksi kelapa sawit, TensorFlow Lite, YOLOv8s-OBB Android mobile application, FFB mass approximation, Palm oil production estimation
Subjects:	Q Science > QA Mathematics > QA336 Artificial Intelligence Q Science > QA Mathematics > QA76.774.A53 Android
Divisions:	Faculty of Intelligent Electrical and Informatics Technology (ELECTICS) > Informatics Engineering > 55201-(S1) Undergraduate Thesis
Depositing User:	Faiq Lidan Baihaqi
Date Deposited:	28 Jul 2026 01:21
Last Modified:	28 Jul 2026 01:21
URI:	http://repository.its.ac.id/id/eprint/138202

Actions (login required)

View Item