Image‐Based Expert‐System Approach to Distress Detection on CRC Pavement

The first step in the successful management of pavements is to locate and identify the distress on all pavements that are candidates for maintenance and rehabilitation. This requires the collection of a large volume of distress data, differentiated by type, extent, and severity. Visual methods of collection have proven to be too labor‐intensive, inconsistent, and hazardous because of exposure to traffic. The need for automated means of data collection being established, currently, videotapes of highway pavement are visually inspected to identify various types of distress. Steps have been taken to analyze videotape images of distress using image‐processing techniques. However, these techniques require a fair amount of human interaction to reach satisfactory results. In this paper, a rule‐based vision system is described that allows the evaluation of concrete distress without the need for any human interaction. The knowledge base of this system contains facts and rules pertaining to prominent features of different types of distress. The reasoning procedure is performed by gathering information on the input image and then by deciding the most effective sequence of image‐processing operations. The system employs the CLIPS environment to achieve easy integration with the image‐processing algorithms written in the C language. The system performance is examined for a large volume of distress image. The results indicate that the system meets all specified requirements, while achieving 85%–90% accuracy of identification at speeds approaching real‐time processing.