فاکتور گیری تانسور برای رده بندی ترک روکش جاده

ABSTRACT Pavement crack classification based on tensor factorization

Infrastructure systems are necessary for supporting society’s functioning as well as economic growth. They are capital intensive assets whose proper functioning is critical to the development of modern societies worldwide. The assets deteriorate over time and as such there is the need for maintenance to ensure they function efficiently during their design life. The most common sign of deterioration in road pavements is cracking. There are various types of cracks namely longitudinal, transverse, alligator and block cracks. Usually, manual surveys are carried out by experienced inspectors who walk along the roadway and note the various types of cracks at different sections of the road. This is a very subjective way of monitoring the condition of the pavement. It is also time consuming and may also pose serious safety risks for the inspectors. In recent times, automated methods which process pavement images for condition monitoring have been developed. Images are taken by cameras attached to specialized vehicles after which processing algorithms are used to determine the types of cracks on the pavement. These algorithms are usually based on neural networks. They are usually associated with problems such as classification errors and long processing times. This paper presents the basic concepts of tensor factorization and also outlines the use of tensor decomposition in pavement crack classification. The two main types of cracks; longitudinal and transverse cracks are considered in this work. The paper has six main parts. The first part is the introduction. This is followed by the second part which describes previous work done in the area of pavement crack classification. The third part focuses on tensors and the basic underlying concepts of tensor factorization. The fourth, fifth and sixth parts are on crack classification based on tensors, results and conclusion respectively.