Sistem Deteksi Api Berdasarkan Domain Ruang Warna YCbCr & HSV Pada Kamera Menggunakan Convolusional Neural Network

Alfikri, Amir Zufar (2025) Sistem Deteksi Api Berdasarkan Domain Ruang Warna YCbCr & HSV Pada Kamera Menggunakan Convolusional Neural Network. Other thesis, Institut Teknologi Sepuluh Nopember.

[thumbnail of 2040211041-Undergraduate_Thesis.pdf] Text
2040211041-Undergraduate_Thesis.pdf - Accepted Version
Restricted to Repository staff only

Download (13MB) | Request a copy

Abstract

Gudang pada lingkungan industri kontraktor memiliki potensi risiko kebakaran yang tinggi akibat aktivitas yang melibatkan bahan mudah terbakar dan peralatan listrik, sehingga memerlukan sistem deteksi dini yang efektif untuk meningkatkan keamanan dan meminimalkan kerugian. Penelitian ini bertujuan merancang dan mengimplementasikan sistem deteksi api menggunakan kamera berbasis segmentasi warna YCbCr dan Hue Saturation Value (HSV) yang diintegrasikan dengan metode Convolutional Neural Network (CNN) untuk meningkatkan akurasi dan kecepatan peringatan dini di gudang kontraktor. Metode penelitian meliputi pengambilan data citra api pada berbagai kondisi pencahayaan menggunakan kamera CCTV, kemudian dilakukan segmentasi warna untuk mengidentifikasi karakteristik visual api dengan menentukan nilai ambang batas optimal pada ruang warna YCbCr dan HSV. Hasil segmentasi warna ini selanjutnya digunakan untuk melatih model CNN You OnlyLook Once (YOLO). Pengujian model CNN dilakukan dengan berbagai variasi parameter pelatihan, termasuk learning rate dan jumlah epoch, untuk mendapatkan performa terbaik. Hasil pengujian segmentasi warna menunjukkan bahwa penyesuaian nilai threshold berhasil mengidentifikasi area api pada berbagai intensitas cahaya. Model CNN YOLOv8n yang dilatih mencapai performa optimal dengan learning rate 0.001, menghasilkan mAP50 sebesar 0,91183 dan mAP50-95 sebesar 0.58404. Evaluasi menggunakan confusion matrix menunjukkan akurasi keseluruhan sistem sebesar 87,5%, dengan kemampuan deteksi kelas "Api" mencapai F1-score 90,91% dan kelas "Bukan Api" dengan F1-score 86,15%. Sistem yang dikembangkan ini menunjukkan potensi yang baik sebagai solusi deteksi api yang akurat dan cepat berbasis analisis visual untuk meningkatkan keamanan di lingkungan industri. ==============================================================================================================================
Warehouses in contractor industrial environments have a high fire risk potential due to activities involving flammable materials and electrical equipment, thus requiring an effective early detection system to enhance security and minimize losses. This research aims to design and implement a camera-based fire detection system using YCbCr and Hue Saturation Value (HSV) color segmentation integrated with the Convolutional Neural Network (CNN) method to improve the accuracy and speed of early warnings in contractor warehouses. The research methodology includes collecting fire image data under various lighting conditions using CCTV cameras, followed by color segmentation to identify visual fire characteristics by determining optimal threshold values in the YCbCr and HSV color spaces. These color segmentation results are then used to train the You Only Look Once (YOLO) CNN model. CNN model testing is conducted with various training parameter variations, including learning rate and number of epochs, to obtain the best performance. The color segmentation test results show that threshold value adjustments successfully identify fire areas at various light intensities. The trained YOLOv8n CNN model achieved optimal performance with a learning rate of 0.001, yielding an mAP50 of 0.91183 and an mAP50-95 of 0.58404. Evaluation using a confusion matrix shows an overall system accuracy of 87.5%, with the Fire class detection achieving an F1-score of 90.91% and the Non-Fire class an F1-score of 86.15%. This developed system demonstrates good potential as an accurate and fast fire detection solution based on visual analysis to enhance security in industrial environments.

Item Type: Thesis (Other)
Uncontrolled Keywords: Deteksi Api, Segmentasi Warna, Pembelajaran Mesin, Early Warning System ============================================================ Fire Detection, Color Segmentation, Machine learning, Early Warning System
Subjects: T Technology > T Technology (General) > T55 Industrial Safety
T Technology > TA Engineering (General). Civil engineering (General) > TA1637 Image processing--Digital techniques. Image analysis--Data processing.
T Technology > TK Electrical engineering. Electronics Nuclear engineering > TK5105.585 TCP/IP (Computer network protocol)
T Technology > TK Electrical engineering. Electronics Nuclear engineering > TK6592.A9 Automatic tracking.
T Technology > TK Electrical engineering. Electronics Nuclear engineering > TK7871.674 Detectors. Sensors
Divisions: Faculty of Vocational > 36304-Automation Electronic Engineering
Depositing User: Amir Zufar Alfikri
Date Deposited: 07 Aug 2025 09:00
Last Modified: 07 Aug 2025 09:01
URI: http://repository.its.ac.id/id/eprint/127966

Actions (login required)

View Item View Item