Deteksi Intrusi Ruangan Otomatis Menggunakan Yolov5 Dan Webcam Dengan Notifikasi Ke Telegram

Sipayung, Jose Bintang Putra (2025) Deteksi Intrusi Ruangan Otomatis Menggunakan Yolov5 Dan Webcam Dengan Notifikasi Ke Telegram. Other thesis, Institut Teknologi Sepuluh Nopember.

Text 07211940000004-Undergraduate_Thesis.pdf - Accepted Version Restricted to Repository staff only Download (6MB) | Request a copy

Abstract

Seiring meningkatnya kebutuhan akan sistem keamanan yang responsif dan cerdas, diperlukan solusi yang mampu mendeteksi intrusi secara otomatis dan memberikan notifikasi secara real-time. Penelitian ini bertujuan untuk merancang dan mengimplementasikan sistem deteksi intrusi ruangan berbasis webcam menggunakan algoritma YOLOv5, serta mengintegrasikannya dengan layanan Telegram sebagai media notifikasi jarak jauh. Model YOLOv5n yang telah dipra-latih pada dataset COCO digunakan untuk mendeteksi keberadaan manusia secara real-time dari umpan video webcam. Hasil pengujian menunjukkan bahwa sistem mampu mendeteksi objek manusia dengan rata-rata akurasi sebesar 89,1%, serta mampu membedakan kondisi keberadaan manusia secara konsisten. Sistem menunjukkan performa real-time dengan kecepatan rata-rata 15 FPS pada resolusi 640×840, menghasilkan waktu deteksi rata-rata 0,07 detik. Akurasi tetap tinggi pada berbagai kondisi nyata, seperti pencahayaan (87,3%), jarak hingga 5 meter (90,1%), sudut kamera (90%), dan variasi pose tubuh, meskipun menurun menjadi 80% saat objek membelakangi kamera. Sistem berhasil mengirim notifikasi otomatis ke Telegram dengan latensi rata-rata 0,3 detik, dilengkapi fitur jeda dan konfirmasi untuk mencegah spam notifikasi. Selain itu, sistem telah diuji pada Raspberry Pi 5 dengan berbagai optimasi untuk memastikan stabilitas dan efisiensi memori selama operasi jangka panjang. Dengan capaian ini, sistem dinilai layak sebagai solusi keamanan ruang berbasis visi komputer yang ringan, responsif, dan dapat diakses dari jarak jauh.
====================================================================================================================================
As the demand for intelligent and responsive security systems continues to grow, there is a need for solutions capable of detecting intrusions automatically and providing real-time notifications. This study aims to design and implement a room intrusion detection system based on a webcam using the YOLOv5 algorithm, integrated with Telegram as a remote notification medium. The YOLOv5n model, pre-trained on the COCO dataset, is employed to detect the presence of humans in real-time from webcam video feeds. Experimental results show that the system can detect human objects with an average accuracy of 89.1%, consistently distinguishing the presence of a person. The system demonstrates real-time performance with an average speed of 15 FPS at a resolution of 640×840, yielding an average detection time of 0.07 seconds. Detection accuracy remains high under various real-world conditions, including lighting variations (87.3%), distances up to 5 meters (90.1%), camera angles (90%), and different body poses, although it drops to 80% when the subject faces away from the camera. The system successfully sends automated Telegram notifications with an average latency of 0.3 seconds, and includes delay and acknowledgment mechanisms to prevent notification spam. Furthermore, the system has been tested on a Raspberry Pi 5, with optimizations to ensure memory efficiency and operational stability over extended periods. With these results, the system is considered a viable lightweight, responsive, and accessible computer vision-based room security solution.

Item Type:	Thesis (Other)
Uncontrolled Keywords:	Computer Vision, Intrusion Detection, Raspberry Pi, Telegram Bot, YOLOv5
Subjects:	T Technology > T Technology (General) T Technology > TK Electrical engineering. Electronics Nuclear engineering T Technology > TK Electrical engineering. Electronics Nuclear engineering > TK7871.674 Detectors. Sensors
Divisions:	Faculty of Intelligent Electrical and Informatics Technology (ELECTICS) > Computer Engineering > 90243-(S1) Undergraduate Thesis
Depositing User:	Jose Bintang Putra Sipayung
Date Deposited:	21 Aug 2025 08:32
Last Modified:	21 Aug 2025 08:32
URI:	http://repository.its.ac.id/id/eprint/126916

Actions (login required)

View Item