Evaluation of Stripping Test Outcomes Using Artificial Neural Networks: A Comparative Analysis of Image Processing and Expert Assessments

Authors : Kadir Akgöl, Ferdi Özbilgin, Ömer Faruk Öztürk, Mehmet Can Tuna, Mücahit Kayiş, Fatih Ulu

Pages : 1647-1667

Doi:10.31466/kfbd.1704724

View : 107 | Download : 133

Publication Date : 2025-12-15

Article Type : Research Paper

Abstract :In this study, the stripping behavior of aggregates used in flexible pavement layers was analyzed by comparing laboratory test results with image processing techniques and expert evaluations. A total of 270 aggregate images obtained under laboratory conditions were analyzed using a custom tool developed in MATLAB. This tool extracted the total stripped area, stripped perimeter, and number of stripped regions for each image and subsequently calculated the stripping percentage. The same specimens were also visually evaluated by laboratory personnel, and expert-based assessments were recorded. The accuracy and consistency of the obtained data were tested using the Artificial Neural Network (ANN) method. The results of this study demonstrate that image processing techniques offer a more objective and reproducible alternative for evaluating stripping behavior. Moreover, the findings highlight the necessity of considering the geometric characteristics of stripped regions in addition to the total stripped area. These insights serve as a guide for improving the accuracy of laboratory stripping tests.
Keywords : Görüntü işleme, Yapay sinir ağları, Soyulma deneyi, Agrega-Bitüm Adezyonu

ORIGINAL ARTICLE URL