Towards a Remote Sensing System for Railroad Bridge Inspections: A Concrete Crack Detection Component

This research paper presents ongoing research work towards developing an effective remote sensing system for bridge inspections. The paper describes the development and evaluation of a five-step prototype image processing algorithm to detect concrete cracks. The algorithm, based on an unsupervised learning approach, extracts pixels that belong to concrete cracks from images. The algorithm was tested using various images collected from field tests. The images included different types of concrete cracks, along with various types of noise factors. The algorithm was successful in detecting the different types of concrete cracks from various images. Contrary to most studies from the literature, these images were not “clean”. That is, the images were collected from field tests and used as inputs to the detection algorithm without any preprocessing activity.