Sistem Keamanan pada Peternakan Sapi Menggunakan Kamera Termal dan Metode YOLO

Nayottama, Hafid Mahdi Anudya (2023) Sistem Keamanan pada Peternakan Sapi Menggunakan Kamera Termal dan Metode YOLO. Other thesis, Institut Teknologi Sepuluh Nopember.

Text 07111840000198-Undergraduate_Thesis.pdf - Accepted Version Restricted to Repository staff only until 1 April 2025. Download (6MB) | Request a copy

Abstract

Peternakan sapi umumnya terdiri dari sapi dalam jumlah banyak, dimasukkan kedalam kandang yang ditata sedemikian rupa sehingga sapi-sapi tidak mungkin dapat melepaskan diri keluar dari kandang tanpa bantuan manusia. Akan tetapi menurut berita-berita, beberapa sapi seperti tersebut acap kali hilang dicuri manusia (maling) ketika diwaktu malam hari. Dengan terbatasnya penjaga dan pengamat diwaktu malam dan siang, suatu cara yg dapat memantau adanya manusia ditempat yang perlu untuk dipantau. Sistem keamanan pada peternakan sapi menggunakan kamera termal sebagai kamera yang dapat digunakan di segala kondisi bahkan pada keadaan gelap gulita. Data gambar yang telah didapatkan oleh kamera termal yang berukuran 160x120 dijadikan dataset yang kemudian dimasukkan ke metode YOLOv7 yang akan membedakan klasifikasi gambar menjadi dua label, yaitu manusia atau sapi. Model pada pelatihan ini memiliki nilai rata-rata presisi sebesar 80%. Sistem keamanan menggunakan speaker dan sistem komunikasi Internet of Things kepada aplikasi android pengguna, sehingga ketika terdeteksi adanya manusia, maka akan dilakukan pemrosesan gambar dan deteksi pada model yang telah dilatih menggunakan mini-pc yaitu Jetson Nano, kemudian gambar dan log persentase deteksi akan di-upload ke internet untuk dikirimkan ke smartphone pengguna serta sistem menyalakan alarm. Sistem deteksi dengan model yang telah dilatih diuji pada kandang sapi saat siang hari, sore hari, malam hari dengan cahaya dan malam hari tanpa cahaya yang memiliki hasil keseluruhan deteksi pada sapi dengan nilai presisi 0.98, recall 0.67, F1-Score 0.79, dan nilai akurasi sebesar 0.79, dan nilai keseluruhan pada deteksi manusia dengan presisi 0.72, recall 0.46, F1-Score 0.57, serta akurasi sebesar 0.75. Sistem keamanan dapat bekerja dengan baik dengan adanya penyusup dan alarm menyala hampir pada seluruh pengujian kecuali pada pengujian malam hari tanpa cahaya jarak dekat yang terjadi false positive pada alarm dengan jangkauan operasi sistem keamanan sebesar 1-8 meter.
==============================================================================================================================
Cattle farming generally consists of many cows, which are put into pens arranged in such a way that it is impossible for the cows to get out of the pen without human assistance. However, according to the news, some cows like these are often stolen by humans (theft) at night. With limited guards and observers at night and day, a way to monitor the presence of humans in places that need to be monitored. The security system on a cattle farm uses a thermal camera as a camera that can be used in all conditions, even in complete darkness. Image data that has been obtained by a thermal camera measuring 160x120 is used as a dataset which is then entered into the YOLOv7 method which will differentiate image classification into two labels, namely humans or cows. The model in this training has an average precision value of 80%. The security system uses speakers and the Internet of Things communication system to the user's android application, so that when a human is detected, image processing and detection will be carried out on a model that has been trained using a mini-pc, namely Jetson Nano, then images and logs of the percentage of detection will be recorded. uploaded to the internet to be sent to the user's smartphone as well as the system to set off an alarm. The detection system with a trained model was tested in a cowshed during the day, evening, night with light and at night without light which has an overall detection result on cattle with a precision value of 0.98, recall 0.67, F1-Score 0.79, and accuracy value of 0.79, and the overall value of human detection with precision 0.72, recall 0.46, F1-Score 0.57, and accuracy of 0.75. The security system can work well in the presence of intruders and the alarm is on in almost all tests except for testing at night without close-range light where there is a false positive on the alarm with the security system operating range of 1-8 meters.

Item Type:	Thesis (Other)
Uncontrolled Keywords:	Jetson Nano, kamera termal, keamanan, YOLOv7
Subjects:	Q Science > Q Science (General) > Q325.5 Machine learning. T Technology > TH Building construction > TH9737 Electronic security systems T Technology > TK Electrical engineering. Electronics Nuclear engineering
Divisions:	Faculty of Intelligent Electrical and Informatics Technology (ELECTICS) > Electrical Engineering > 20201-(S1) Undergraduate Thesis
Depositing User:	Hafid Mahdi Anudya Nayottama
Date Deposited:	03 Feb 2023 04:27
Last Modified:	03 Feb 2023 04:31
URI:	http://repository.its.ac.id/id/eprint/96062

Actions (login required)

View Item