Automated Interpretation and Assessment of Multi-Sensory Underground Infrastructure Data

摘要

Traditionally, methods used to assess status and inventory condition of underground infrastructure havernbeen based on after-the-fact information. Often, these infrastructure assets get neglected until they sufferrncatastrophic failures, which are inconvenient and costly to repair. To determine the health of infrastructurernsystems, regular and accurate assessment is essential. Many technologies have evolved over the past fivernyears for inspecting, monitoring, and evaluating infrastructure problems. However, the effective application ofrnthese technologies has been slow, and in many cases, very reluctant. This problem can be addressed in twornsteps: First, the accuracies and precisions of these infrastructure inspection technologies must be analyzedrnin order to quantify the variances of various technologies. Then the interpretation of the data provided byrnthese technologies needs to be improved, both in terms of speed and accuracy.rnRecent advances in optical sensors and computing technologies have led to the development of inspectionrnsystems for underground facilities such as water lines, sewer pipes and telecommunication conduits. It isrnnow possible for inspection technologies that require no human entry into underground structures, to be fullyrnautomated from data acquisition to data analysis, and eventually to condition assessment. This paperrndescribes the development of an automated data interpretation system for sanitary sewer pipelines. Thernproposed system utilizes artificial neural networks to recognize various types of defects in sanitary sewerrnpipelines. The framework of this system includes modification of digital images for preprocessing, imagernfeature segmentation, utilization of multi-neural networks for feature pattern recognition, and the fusion ofrnmultiple neural networks via the use of fuzzy logic systems.