Optimasi deteksi retakan jalan menggunakan filter sobel dan klasifikasi Gaussian Naïve Bayes

Hidayat, Fakhar Muhammad, Crysdian, Cahyo ORCID: https://orcid.org/0000-0002-7488-6217 and Lestari, Tri Mukti ORCID: https://orcid.org/0009-0005-9416-7905 (2026) Optimasi deteksi retakan jalan menggunakan filter sobel dan klasifikasi Gaussian Naïve Bayes. JISKA : Jurnal Informatika Sunan Kalijaga, 11 (2). pp. 155-168. ISSN 2528-0074

Preview

Text 27800.pdf - Published Version Available under License Creative Commons Attribution Non-commercial. Download (1MB) | Preview

Abstract

ENGLISH

Manual identification of road damage using simple measuring tools is considered inefficient, subjective, and time-consuming, hindering the infrastructure repair process. This study aims to optimize automatic road crack detection by combining edge detection for feature extraction and Gaussian Naive Bayes (GNB) classification. This research utilizes the Road Surface Classification Dataset (RSCD), consisting of 1000 concrete road images with balanced class proportions. The research process includes image acquisition, segmentation, and preprocessing using the Sobel filter to extract edge features and erosion to refine crack representation. Statistical features in the form of black pixel count and edge length are extracted as model inputs. Experiments were conducted using three data split scenarios (70:30, 80:20, 90:10) validated with the K-Fold Cross Validation method. The test results show that the 90:10 data split scenario yields the most optimal and stable performance, achieving 88% accuracy, 91.67% precision, 84.62% recall, and an F1-Score of 88%. This study optimises the balance between computational efficiency and detection accuracy through a lightweight hybrid approach that integrates edgebased feature extraction with probabilistic classification.

INDONESIA

Identifikasi kerusakan jalan secara manual menggunakan alat ukur sederhana dinilai kurang efisien, subjektif, dan membutuhkan waktu lama sehingga dapat menghambat proses perbaikan infrastruktur. Penelitian ini bertujuan untuk mengoptimalkan deteksi retakan jalan secara otomatis dengan menggabungkan deteksi tepi untuk ekstraksi fitur dan klasifikasi menggunakan Gaussian Naive Bayes (GNB). Penelitian ini memanfaatkan Road Surface Classification Dataset (RSCD) yang terdiri dari 1000 citra jalan beton dengan proporsi kelas yang seimbang. Proses penelitian meliputi akuisisi citra, segmentasi, preprocessing menggunakan filter Sobel untuk mengekstraksi fitur tepi dan erosi untuk menyempurnakan representasi retakan. Fitur statistik berupa jumlah piksel hitam dan panjang tepi diekstraksi sebagai input untuk model. Uji coba dilakukan menggunakan tiga skenario pembagian data (70:30, 80:20, 90:10) yang divalidasi dengan metode K-Fold Cross Validation. Hasil pengujian menunjukkan bahwa skenario pembagian data 90:10 memberikan performa paling optimal dan stabil dengan akurasi mencapai 88%, presisi 91,67%, recall 84,62%, dan F1-Score 88%. Penelitian ini mengoptimalkan keseimbangan antara efisiensi komputasi dan akurasi deteksi melalui pendekatan hibrida ringan yang mengintegrasikan ekstraksi fitur berbasis deteksi tepi dengan klasifikasi probabilistik.

Item Type:	Journal Article
Keywords:	road crack detection; Gaussian Naïve Bayes; sobel filter; image classification; image processing; deteksi retakan jalan; Gaussian Naïve Bayes; filter sobel; klasifikasi citra; pengolahan citra
Subjects:	08 INFORMATION AND COMPUTING SCIENCES > 0801 Artificial Intelligence and Image Processing > 080104 Computer Vision 08 INFORMATION AND COMPUTING SCIENCES > 0801 Artificial Intelligence and Image Processing > 080106 Image Processing 08 INFORMATION AND COMPUTING SCIENCES > 0801 Artificial Intelligence and Image Processing > 080199 Artificial Intelligence and Image Processing not elsewhere classified 08 INFORMATION AND COMPUTING SCIENCES > 0801 Artificial Intelligence and Image Processing
Divisions:	Faculty of Technology > Department of Informatics Engineering
Depositing User:	Dr. Cahyo Crysdian
Date Deposited:	13 Jul 2026 09:20

Downloads

Origin of downloads

Actions (login required)

View Item