Asphalt pavement segregation detection method based on LBP-GLCM

Xun Zhao, Feiyun Xu

Abstract: Asphalt pavement segregation is a common failure that leads to a decline in road life and safety. In order to solve the problem of low accuracy, a segregation detection method based on texture feature extraction and support vector machine (SVM) classifier is proposed. The study first manually marked 300 asphalt pavement images. Then, the original image is encoded by Local Binary Pattern (LBP). Next, four texture features (correlation, contract, energy, homogeneity) are calculated through the gray level co-occurrence matrix (GLCM). Later, input the texture feature matrix into SVM classifier to train the classifier. Finally, the classification results show that the accuracy of the SVM classifier is 85% compared with the results of manual labeling. The results of this study can potentially be used for large-scale assessment of asphalt segregation.

External IDs:dblp:conf/indin/ZhaoX20