WP3 - Computer Vision for Real Time Defect Detection and Assessment
The objective of this WP is to adapt the existing FIAS cognitive vision platform and apply it to bridge inspection. This platform has a scalable cognitive architecture that is motivated from systems engineering and brain science principles. We will use our strengths in systematic engineering of vision systems to translate application requirements to specific hypotheses generators (i.e. quick look mechanisms that flag areas for finer examination) and use finer inspection strategies to estimate crack width and perform 3D measurements. This WP will also address the development of a module for self-localisation from vision, based on 3D geometric models of global bridge layout plans available and input from other geo-location sensors from the robot. The global position and orientation of the vision system with respect to the world is provided as output for other sensor fusion and navigation modules. The cognitive vision system will use machine learning methods to classify cracks and will construct on-the-fly instance specific models of cracks, measure distances between parallel cracks, identify swollen and spalled areas, take high resolution photographs of the bearings to inspect for level of deformation.
Work Package Leader: