Efficient Pavement Crack Area Classification Using Gaussian Mixture Model Based Features

Pavement cracks are caused by various factors such as aged deterioration, load and weather conditions, and so on. As these cracks reduce the safety of road traffic, regular inspections are necessary. In recent years, various crack detection methods using pavement images have been proposed. However, those often have problems with accuracy and processing time. Therefore, in order to reduce the amount of calculation, we devised an efficient method to narrow down the area containing cracks in the pavement image. The method consists of crack feature extraction combining Gaussian mixture model and image filtering, and classification by support vector machine. The experimental results show that our proposed method is the most efficient in accuracy and processing speed compared with conventional methods.