Novel Image Database Analysis System for Maintenance of Transportation Facility: Phase I

摘要

Transportation is critical to the global economy and plays a particularly vital role in this regions economic growth. Transportation infrastructures such as highways, streets, and bridges represent one of the largest public investments of many governments. Maintaining and managing the vast and aging transportation infrastructures with limited resources is a most challenging task for nearly all transportation agencies. Monitoring and evaluating the physical condition of the transportation infrastructure and scheduling for timely repair are essential for effective infrastructure management. However, human inspection can be time consuming and prone to inconsistent results due to differences in judgments. Digital imaging technology has been proposed as a viable alternative to human inspection and provides automated inspection, monitoring, and pattern recognition, as the cost of imaging technologies has become more affordable. The objective of this research was to assist transportation agencies efficiently record, monitor and evaluate the conditions of transportation infrastructure assets so as to more effectively managing the needs for maintenance or rehabilitation with minimum total costs and least interruptions of services. The researchers reviewed relevant literatures and designed the framework of a prototype imaging system to systematically and automatically extract features from digital images of pavement surfaces. The focus was on processing and transforming the roadway images for analysis, and extracting visible cracking data including crack length, width, and patterns from the processed roadway images. A new crack connectivity algorithm to rebuild the crack pattern after noise reduction was developed and implemented. This study was jointly funded by UT-UTC and MIOH-UTC. Phase II of this project, which continues the development of the prototype and focuses on improved quantification of cracking in terms of its length and width, and uses wavelet transformations to extract and identify different crack patterns in order to match the imaging processing results with human ratings, was subsequently funded by both sponsors.