Dynamic Pavement Delineation and Visualization Approach Using Data Mining

Abstract

Highway agencies have been using automated and semiautomated data collection methods such as laser scanning and ultrasonic waves, resulting in the collection of an enormous amount of high-density pavement condition data. Most agencies are now able to quantify the extent and severity of distresses for extremely short lengths of pavement sections. A scientific and dynamic method to aggregate small pavement sections into reasonably sized segments plays an important role in implementing several pavement management tasks. This paper proposes a new delineation method for pavement sections that finds homogenous segments by considering multiple pavement distresses using affinity propagation clustering. A case study was conducted using pavement condition data in Iowa to illustrate the capabilities and applications of the proposed segmentation framework. The results of the case study showed that agencies can evaluate the accuracy of delineated segments by changing the delineation parameters, including minimum segment length. The proposed algorithm is expected to significantly enhance many pavement management applications such as deterioration modeling and maintenance programming.