A method to evaluate the segregation of compacted asphalt pavement by processing the images of paved asphalt mixture. (10th November 2019)

• Record Type: Journal Article
• Authors: Cong, Lin; Shi, Jiachen; Wang, Tongjing; Yang, Fan; Zhu, Tiantong
• Is Part Of: Construction & building materials. Volume 224(2019)
• Page Start: 622
• Page End: 629
• Publication Date: 2019-11-10
• Subjects: Asphalt pavement -- Segregation -- Image processing -- Gray level co-occurrence matrix -- Principal component analysis -- Naive Bayesian classifier; Building materials -- Periodicals; 624.18
• DOI: 10.1016/j.conbuildmat.2019.07.041
• Languages: English
• ISSNs: 0950-0618

Abstract

Highlights: There is a link between the spatial distribution of aggregates in the paved mixture and the compacted pavement. The texture features of the image were calculated to characterize the segregation. The Bayesian classifier is used to classify whether the segregation occurs. This method can prejudge the segregation of the compacted pavement using the image of the paved mixture. Abstract: Segregation in hot-mix asphalt pavement is a common failure during the construction process. The prevailing segregation detection methods can be used to detect and evaluate segregation only after segregation occurs. This study proposes a real time segregation detection method by using machine learning classifier to categorize the images of the paved mixture (IPM) during construction. The study first manually labeled 224 various levels of hot mix asphalt segregation images. Then, 14 texture features such as contrast, correlation of the IPM were calculated by the gray level co-occurrence matrix (GLCM). Next, the principal component analysis (PCA) was done to reduce the 14 features to 6 main components. Later on, the 6 main components were fed to a Naive Bayesian classifier to categorize the segregation level. Finally, the classification results indicate that the Naïve Bayesian classifier has 80% accuracy when compared with the manually labelled results. Results of this study can potentially be adapted for real-time and large-scale hot mix asphalt segregation evaluation.