Muhammad Eka, Setio Aji (2023) ANALISIS PERFORMA CONVOLUTIONAL NEURAL NETWORK UNTUK PENGENALAN WAJAH PADA SPEKTRUM INFRAMERAH. S1 thesis, UNIVERSITAS SULTAN AGENG TIRTAYASA.

	Text (FULL TEXT SKRIPSI) Muhammad Eka Setio Aji_3332190068_Fulltext.pdf Restricted to Registered users only Download (2MB) | Request a copy
	Text (COVER BAB 1) Muhamamd Eka Setio Aji_3332190068_01.pdf Restricted to Registered users only Download (1MB) | Request a copy
	Text (BAB 2) Muhammad Eka Setio Aji_3332190068_02.pdf Restricted to Registered users only Download (628kB) | Request a copy
	Text (BAB 3) Muhammad Eka Setio Aji_3332190068_03.pdf Restricted to Registered users only Download (619kB) | Request a copy
	Text (BAB 4) Muhammad Eka Setio Aji_3332190068_04.pdf Restricted to Registered users only Download (988kB) | Request a copy
	Text (BAB 5) Muhammad Eka Setio Aji_3332190068_05.pdf Restricted to Registered users only Download (352kB) | Request a copy
	Text (REFRENSI) Muhammad Eka Setio Aji_3332190068_Ref.pdf Restricted to Registered users only Download (404kB) | Request a copy
	Text (LAMPIRAN) Muhammad Eka Setio Aji_3332190068_Lamp.pdf Restricted to Registered users only Download (4MB) | Request a copy
	Text (CEK PLAGIASI) Muhammad Eka Setio Aji_3332190068_CP.pdf Restricted to Registered users only Download (18MB) | Request a copy

Abstract

Infrared images have the potential without being affected by light illumination. Several drawbacks, such as less information and low texture quality, can be solved by using methods with the potential for extracting the feature in a multi-layer. CNN with Haar are solution to solve this problem. Accuracy obtained was 0.98, and the average accuracy score was 0.94. This method has the potential for face recognition in infrared images. Data were acquired in 2 modalities simultaneously in infrared and RGB images. Data will be compared in infrared and RGB to obtain the best performance for face recognition in low illumination was 9 to 21 Lux. The result obtained from the comparison showed the infrared images were not affected by low illumination. At the same time, RGB images were affected by illumination, and it will affect to the reliability of the system.

Item Type:	Thesis (S1)
Contributors:	Contribution Contributors NIP/NIM Thesis advisor ROCKY, ALFANZ 0028038103
Additional Information:	Citra inframerah memiliki keunggulan dengan tidak terpengaruh oleh iluminasi cahaya jika dibandingkan dengan kamera RGB biasa. Disamping kelemahan dari citra inframerah yang rendah informasi dan kualitas tekstur, hal tersebut dapat diatasi dengan menggunakan metode yang memiliki potensi untuk melakukan ekstraksi fitur dengan memanfaatkan kedalaman layar. CNN dengan Haar Classifier dapat digunakan menjadi solusi untuk mengatasi permasalahan tersebut. Dengan tingkat akurasi tertinggi hingga mencapai 0,98 dan rata-rata sebesar 0,94 metode ini memiliki kehandalan yang tinggi untuk merekognisi subjek. Pengambilan data dengan 2 modalitas sekaligus dalam citra infra-merah dan RGB. Komparasi dilakukan dengan tujuan mengetahui perbandingan performa proses pengenalan wajah dengan kondisi lingkungan rendah pencahayaan antara 9 hingga 21 Lux. Hasil komparasi menunjukan kehandalan dari citra infra-merah tidak terpengaruh dengan iluminasi cahaya, sementara citra RGB sangat terpengaruh dengan iluminasi cahaya dan hal tersebut sangat mempengaruhi kehandalan dari sistem.
Subjects:	Q Science > QA Mathematics > QA75 Electronic computers. Computer science Q Science > QA Mathematics > QA76 Computer software
Divisions:	03-Fakultas Teknik 03-Fakultas Teknik > 20201-Jurusan Teknik Elektro
Depositing User:	Mr Muhammad Eka Setio Aji
Date Deposited:	19 Jan 2023 12:08
Last Modified:	19 Jan 2023 12:08
URI:	http://eprints.untirta.ac.id/id/eprint/20023

Available Versions of this Item

• ANALISIS PERFORMA CONVOLUTIONAL NEURAL NETWORK UNTUK PENGENALAN WAJAH PADA SPEKTRUM INFRAMERAH. (deposited UNSPECIFIED)
  • ANALISIS PERFORMA CONVOLUTIONAL NEURAL NETWORK UNTUK PENGENALAN WAJAH PADA SPEKTRUM INFRAMERAH. (deposited 19 Jan 2023 12:08) [Currently Displayed]

Actions (login required)

View Item