Desain Sistem Deteksi Dan Klasifikasi Permukaan Datar Secara Non-kontak Berbasis Convolusional Neural Network

Mustaqim, Bagus (2026) Desain Sistem Deteksi Dan Klasifikasi Permukaan Datar Secara Non-kontak Berbasis Convolusional Neural Network. Other thesis, Institut Teknologi Sepuluh Nopember.

Text 5001221060-Undergraduate_Thesis.pdf - Accepted Version Restricted to Repository staff only Download (4MB) | Request a copy

Abstract

Kerusakan pada permukaan beton merupakan salah satu bentuk kerusakan pada struktur yang dapat menurunkan kualitas, kekuatan, dan umur infrastruktur. Apabila kerusakan tidak terdeteksi sejak dini, kerusakan tersebut dapat berkembang menjadi lebih parah. Oleh karena itu, diperlukan suatu metode deteksi yang mampu mendukung pemantauan kondisi permukaan secara berkala sehingga kerusakan dapat diidentifikasi sejak awal. Metode deteksi berbasis pengolahan citra menggunakan kamera telah banyak digunakan, namun memiliki keterbatasan karena sangat bergantung pada kondisi pencahayaan dan sudut pengambilan gambar. Untuk mengatasi keterbatasan tersebut, penelitian ini mengembangkan prototype sistem deteksi klasifikasi permukaan datar secara non-kontak berbasis sensor Time of Flight (ToF) dan algoritma Convolutional Neural Network (CNN). Sistem dirancang menggunakan empat sensor VL53L0X yang disusun secara sejajar dan diintegrasikan dengan microcontroller ESP32 untuk memperoleh data profil kedalaman permukaan datar. Data hasil pemindaian kemudian diolah menjadi set data dan digunakan untuk melatih model CNN dalam membedakan kondisi permukaan datar utuh dan berlubang. Sistem dilengkapi antarmuka berbasis web untuk mempermudah pengoperasian sistem dan menampilkan hasil klasifikasi secara real-time. Model CNN dilatih menggunakan data dari berbagai variasi permukaan bata ringan yang merepresentasikan dua kategori, yaitu utuh dan berlubang. Selanjutnya, model CNN diuji menggunakan set data yang berbeda dari data pelatihan. Hasil pengujian menunjukkan bahwa kombinasi sensor ToF dan CNN efektif untuk mendeteksi dan mengklasifikasikan kondisi permukaan datar secara non-kontak
=======================================================================================================================================
Damage to concrete surfaces is one form of structural deterioration that can reduce the quality, strength, and service life of infrastructure. If not detected at an early stage, such damage may progress and become more severe. Therefore, an effective detection method is required to support periodic surface condition monitoring so that damage can be identified at an early stage. Camera-based image processing methods have been widely used for surface defect detection. However, their performance is highly dependent on lighting condition and image acquisition angles. To overcome these limitations, this study develops a non-contact flat surface classification detection system based on Time of Flight (ToF) sensors and a Convolutional Neural Network (CNN) algorithm. The proposed system employs four VL53L0X sensors arranged in parallel and integrated with ESP32 microcontroller to acquire flat surface depth profile data. The scanned data are then processed into a dataset and used to train a CNN model to distinguish between intact and damaged flat surfaces containing potholes. The system is equipped with a web-based interface to facilitate system operation and display classification results in real-time. The CNN model is trained using data obtained from various Autoclaved Aerated Concrete (AAC) surface conditions representing two categories, namely intact and potholes. Subsequently, the model is evaluated using a testing dataset that is different from the training dataset. The experiment results demonstrate that the combination of ToF sensors and CNN is effective for detecting and classifying flat surface conditions in a non-contact manner.

Item Type:	Thesis (Other)
Uncontrolled Keywords:	Deteksi Non-Kontak, Klasifikasi Permukaan Datar, Non-Destructive Testing (NDT), Time of Flight (ToF), Convolutional Neural Network (CNN), Flat Surface Classification, Non-Contact Inspection
Subjects:	T Technology > T Technology (General)
Divisions:	Faculty of Science and Data Analytics (SCIENTICS) > Physics > 45201-(S1) Undergraduate Thesis
Depositing User:	Bagus Mustaqim
Date Deposited:	21 Jul 2026 01:54
Last Modified:	21 Jul 2026 01:54
URI:	http://repository.its.ac.id/id/eprint/135975

Actions (login required)

View Item