Sistem Estimasi Volume Pupuk Berbasis Point Cloud Menggunakan Metode Stereo Vision Statis

Aldianto, Mochamad Dani (2026) Sistem Estimasi Volume Pupuk Berbasis Point Cloud Menggunakan Metode Stereo Vision Statis. Other thesis, Institut Teknologi Sepuluh Nopember.

Text 2040221028-Undergraduate_Thesis.pdf - Accepted Version Restricted to Repository staff only Download (5MB) | Request a copy

Abstract

Industri pupuk, seperti PT Petrokimia Gresik, membutuhkan sistem pengawasan aliran material yang sangat presisi pada jalur konveyor. Penggunaan sensor mekanis konvensional berupa load cell terbukti rentan terhadap keausan fisik, getaran konveyor, dan interferensi elektromagnetik (EMI) yang memicu degradasi akurasi yang fatal. Penelitian ini bertujuan untuk merancang sistem estimasi volume pupuk nirkontak (non-contact) berbasis Point Cloud tiga dimensi menggunakan metode Stereo Vision statis sebagai alternatif pengukuran yang tahan banting. Akuisisi citra dilakukan menggunakan sepasang kamera stereo global shutter, yang kemudian diproses melalui tahapan kalibrasi, rektifikasi Bouguet, pencocokan disparitas Semi-Global Block Matching (SGBM), ekstraksi bidang referensi konveyor menggunakan algoritma RANSAC, dan perhitungan volume integrasi luasan Elevation Grid 2.5D berskala resolusi 1×1 mm. Hasil pengujian sistem menunjukkan tingkat akurasi yang sangat optimal pada pengujian material pupuk bervolume 1000 mL dengan pencapaian akurasi 98,23% (galat -1,77%). Namun, pada pengujian volume 500 mL, sistem mencatatkan penyusutan ekstrem (under-estimate) dengan akurasi 77,71% (galat -22,29%) sebagai akibat dari pemotongan elevasi dasar (Height Threshold 15 mm) pada area tepian pupuk yang melandai. Penelitian ini membuktikan bahwa metode penginderaan spasial secara statis sangat layak digunakan sebagai landasan (proof of concept) pengganti sensor mekanis pada objek bervolume padat, meskipun masih menuntut kompensasi perangkat lunak tingkat lanjut untuk menangani objek yang berprofil landai.
===================================================================================================================================
The fertilizer industry, such as PT Petrokimia Gresik, requires highly precise material flow monitoring systems on conveyor lines. The use of conventional mechanical sensors like load cells has proven vulnerable to physical wear, conveyor vibration, and electromagnetic interference (EMI), leading to fatal accuracy degradation. This study aims to design a noncontact, 3D Point Cloud-based fertilizer volume estimation system using a static Stereo Vision method as a robust alternative measurement tool. Image acquisition was performed using a pair of global shutter stereo cameras, subsequently processed through calibration, Bouguet rectification, Semi-Global Block Matching (SGBM) disparity extraction, conveyor reference plane extraction using the RANSAC algorithm, and volume calculation via a 1×1 mm resolution 2.5D Elevation Grid integration. The system's test results demonstrated optimal accuracy in the 1000 mL fertilizer volume test, achieving a 98.23% accuracy rate (-1.77% error). However, in the 500 mL test, the system recorded an extreme under-estimate with 77.71% accuracy (22.29% error) as a direct consequence of base elevation filtering (15 mm Height Threshold) cutting off the shallow edges of the fertilizer. This study scientifically proves that static spatial sensing is highly feasible as a proof of concept to replace mechanical sensors for solid-volume objects, although it still requires advanced software compensation to handle shallow-profile objects.

Item Type:	Thesis (Other)
Uncontrolled Keywords:	Stereo Vision, Estimasi Volume, SGBM, Point Cloud, Pengukuran NonContact, Stereo Vision, Volume Estimation, SGBM, Point Cloud, Non-Contact Measurement.
Subjects:	T Technology > TK Electrical engineering. Electronics Nuclear engineering > TK7871.674 Detectors. Sensors T Technology > TK Electrical engineering. Electronics Nuclear engineering > TK7878 Electronic instruments T Technology > TK Electrical engineering. Electronics Nuclear engineering > TK7882.P3 Pattern recognition systems
Divisions:	Faculty of Vocational > 36304-Automation Electronic Engineering
Depositing User:	Mochamad Dani Aldianto
Date Deposited:	20 Aug 2026 08:31
Last Modified:	20 Aug 2026 08:31
URI:	http://repository.its.ac.id/id/eprint/144373

Actions (login required)

View Item