Bilal, Endroen Muhammad (2024) Klasifikasi Vokalisasi Songbird (Genus Lonchura sp.) Menggunakan Convolutional Neural Network. Other thesis, Institut Teknologi Sepuluh Nopember.
Text
BukuTA_5009201183_Endroen Muhammad Bilal.pdf - Accepted Version Restricted to Repository staff only Download (2MB) | Request a copy |
Abstract
Indonesia merupakan negara yang kaya akan keanekaragaman biodiversitas. Burung, khususnya Burung pipit (genus lonchura sp.), merupakan salah satu jenis hewan yang sering dipantau habitatnya karena populasi burung dapat mencerminkan kesehatan ekosistem. Namun, Burung pipit sering menjadi ancaman bagi petani padi sawah dan sangat sulit untuk dikendalikan, menyebabkan kerugian yang signifikan akibat gagal panen. Pemantauan burung di alam liar sulit dilakukan karena penyebaran burung yang luas dan kecepatan terbang mereka yang tinggi. Oleh karena itu, identifikasi suara atau kicauan burung merupakan solusi efektif untuk memantau kehadiran Burung pipit di lahan pertanian agar dapat dilakukan pengembangan strategi mitigasi yang efisien. Penelitian ini menggunakan Convolutional Neural Network (CNN) untuk menganalisis suara burung menggunakan citra spectrogram dengan fokus pada kelas-kelas vokalisasi utama (Calls, Songs & Background), dengan dataset dari rekaman langsung dan dataset online dari laman xeno-canto sebagai data validasi. Proses training dan validasi dengan CNN sederhana menghasilkan akurasi mencapai 99,56%. Metode ini diharapkan dapat memberikan kontribusi signifikan dalam pengembangan sistem pemantauan dan pengendalian Burung pipit untuk pelestarian lingkungan.
============================================================
Indonesia is a country rich in biodiversity. Birds, especially the songbirds (Genus lonchura sp.), are closely monitored as their populations reflect ecosystem health. However, songbirds (Genus lonchura sp.) often threaten rice farmers and are difficult to control, leading to significant losses due to crop failures. Monitoring birds in the wild is challenging due to their wide distribution and high flight speeds. Therefore, identifying bird calls or songs is an effective solution to monitor songbirds (Genus lonchura sp.) presence in agricultural lands and develop efficient mitigation strategies. This study employs Convolutional Neural Network (CNN) to analyze bird sounds using spectrogram images, focusing on primary vocalization classes (Calls, Songs & Background), with datasets from direct recordings and online sources like xeno-canto for validation. Training and validation using a simple CNN achieved an accuracy of 99.56%. This method is expected to significantly contribute to developing monitoring and control systems for songbirds (Genus lonchura sp.), thereby contributing to environmental conservation.
Item Type: | Thesis (Other) |
---|---|
Uncontrolled Keywords: | Burung Pipit, Convolutional Neural Network, Spektrogram, Citra, Confusion matrix. Lonchura Sp. Songbirds, Convolutional Neural Network, Spectrogram, image, Confusion matrix |
Subjects: | T Technology > TK Electrical engineering. Electronics Nuclear engineering > TK7882.P3 Pattern recognition systems |
Divisions: | Faculty of Industrial Technology > Physics Engineering > 30201-(S1) Undergraduate Thesis |
Depositing User: | Endroen Muhammad Bilal |
Date Deposited: | 04 Aug 2024 10:00 |
Last Modified: | 04 Aug 2024 10:00 |
URI: | http://repository.its.ac.id/id/eprint/110012 |
Actions (login required)
View Item |