Sistem Perangkap Hewan Menggunakan Kamera Termal dan Neural Network

Simamora, Misael Joy Edbert Araputra (2026) Sistem Perangkap Hewan Menggunakan Kamera Termal dan Neural Network. Other thesis, Institut Teknologi Sepuluh Nopember.

Text 07111940000129-Undergraduate_Thesis.pdf Restricted to Repository staff only Download (9MB) | Request a copy

Abstract

Produktivitas tanaman padi di Indonesia menghadapi tantangan serius akibat serangan hama tikus, sementara metode pengendalian konvensional seperti rodentisida memiliki risiko terhadap kesehatan dan lingkungan. Penelitian ini mengusulkan sistem identifikasi tikus berbasis kecerdasan buatan menggunakan kamera termal MLX90640 beresolusi rendah (32×24 piksel) sebagai alternatif yang lebih aman dan selektif. Kamera termal dipilih karena mampu mendeteksi objek berdasarkan distribusi suhu, sehingga tetap efektif pada kondisi pencahayaan rendah. Hamster digunakan sebagai hewan uji untuk merepresentasikan tikus dengan mempertimbangkan kesetaraan karakteristik termal. Sistem prototipe direalisasikan menggunakan Raspberry Pi 5 dan diuji pada lingkungan indoor dengan tiga kelas kondisi, yaitu lingkungan tanpa objek, keberadaan objek non-tikus, dan keberadaan hamster. Evaluasi dilakukan dengan membandingkan beberapa arsitektur jaringan saraf, meliputi simple neural network, deep neural network, serta convolutional neural network (CNN). Hasil pengujian menunjukkan bahwa prototipe berhasil berfungsi sesuai rancangan dan mampu mendeteksi keberadaan hewan uji secara andal. Meskipun beresolusi rendah, citra termal MLX90640 terbukti cukup informatif untuk tugas klasifikasi, dengan model CNN dasar mencapai akurasi sekitar 88,95%, sementara beberapa model lainnya menunjukkan performa mendekati 100% pada kondisi indoor. Hasil ini menegaskan kelayakan kamera termal beresolusi rendah untuk sistem perangkap tikus cerdas yang ekonomis dan aplikatif.
=================================================================================================================================
Rice productivity in Indonesia faces significant challenges due to rodent infestations, while conventional control methods such as chemical rodenticides pose risks to human health and the environment. This study proposes intelligent rodent identification system utilizing artificial intelligence and a low-resolution thermal camera, MLX90640 (32×24 pixels), as a safer and more selective alternative. Thermal imaging is employed due to its ability to detect objects based on temperature distribution, enabling reliable operation under low-light or dark conditions. Hamsters are used as experimental subjects to represent rodents, considering their comparable thermal characteristics and ethical constraints. A prototype system was implemented using a Raspberry Pi 5 and evaluated in an indoor environment with three classification classes: empty environment, non-rodent object presence, and rodent presence. Several neural network architectures, including simple neural networks, deep neural networks, and convolutional neural networks (CNNs), were evaluated. Experimental results demonstrate that the proposed prototype operates as intended and can reliably detect the presence of the test animal. Despite its low spatial resolution, the MLX90640 thermal images contain sufficient temperature information for classification tasks. The basic CNN model achieved an accuracy of approximately 88.95%, while several other models reached performance close to 100% under indoor conditions. These findings confirm the feasibility of low-resolution thermal cameras for cost-effective and practical intelligent rodent trap systems.

Item Type:	Thesis (Other)
Uncontrolled Keywords:	Kamera Termal, Neural network, Raspberry Pi 5 Thermal Camera, Neural network, Raspberry Pi 5
Subjects:	T Technology > TK Electrical engineering. Electronics Nuclear engineering T Technology > TK Electrical engineering. Electronics Nuclear engineering > TK7882.P3 Pattern recognition systems
Divisions:	Faculty of Intelligent Electrical and Informatics Technology (ELECTICS) > Electrical Engineering > 20201-(S1) Undergraduate Thesis
Depositing User:	Misael Joy Edbert Araputra Simamora
Date Deposited:	30 Jan 2026 01:57
Last Modified:	30 Jan 2026 01:57
URI:	http://repository.its.ac.id/id/eprint/131068

Actions (login required)

View Item