An Automated Computer Vision System for Damage Detection and Monitoring in Tunnel Inspection

Add to your list(s) Download to your calendar using vCal

• Krisada Chaiyasarn, CUED
• Friday 04 March 2011, 16:00-17:00
• Engineering Department - Lecture Room 6.

If you have a question about this talk, please contact Anama Lowday.

One of the greatest challenges facing civil engineers in the 21st century is the stewardship of aging civil engineering infrastructure. Maintainance and repair works are regularly carried out to ensure safety and serviceability of these structures. One common practice in such maintenance work is visual inspection in order to detect and to monitor anomalies (e.g. cracks, spalling and staining) on the tunnel surface. This routine inspection work has a number of drawbacks, including, labour-intensive, costly, and inaccurate. In this project, a prototype of an automated damage detection system based on state-of-the-art computer vision technology was develped to assist the inspection work. The system is composed of two parts: (1) a novel method in inspection reporting; and (2) automatic detection of changes from images.

The first part of the system created a mosaic image from a large image database taken from real underground environment. The system is composed of two main components: a Structure from Motion (SfM) system; and the 3D surface fitting of a point cloud. These components enable our system to mosaic images obtained from free camera motion as well as to cope with a complex geometry of the scene, unlike most standard image stitching software packages that fail to generate good mosaics because they rely on a number of strong assumptions on the camera motion or on the scene geometry. The main advantage of our system is that it can create a large mosaic image of the tunnel surface with little distortion. Subsequently, it can be used as an inspection report.

The second part of the system automatically performs image analysis by change detection, which is an approach to monitor structural changes from images taken at different times. Although, there have been some promising change detection algorithms in the civil engineering literatures; all of them assume images are taken at the same camera positions. On the contrary, our system can perform change detection on images obtained from arbitrary camera positions, which was made possible by accurate registration from the SfM system. The output from the first part, i.e. a mosaic image, is used as a database in which a newly acquired image will be compared in order to detect changes on the tunnel surface.

In this talk, the framework of our system will be explained. The contributions made in each component in our system pipeline will be highlighted and some key results will be presented.

This talk is part of the Engineering Department Geotechnical Research Seminars series.

This talk is included in these lists:

• Alex
• All Talks (aka the CURE list)
• Cambridge University Engineering Department Talks
• Cambridge talks
• Centre for Smart Infrastructure & Construction
• Civil Engineering Talks
• Computational Continuum Mechanics Group Seminars
• Engineering Department - Lecture Room 6
• Engineering Department Geotechnical Research Seminars
• Featured lists
• Interested Talks
• School of Technology
• Trust & Technology Initiative - interesting events
• bld31

Note that ex-directory lists are not shown.