Self-correcting neural network in road pavement diagnostics

The paper focuses on application of the self-correcting neural network in the process of road pavement diagnostics. It provides a discussion on a method of road inspection based on the proposed neural network solution and a measuring station which uses stereo vision of road pavement. The solution proposed was verified in real-life conditions, i.e. in a road with different types of pavement defects. With reference to a comparison of the results, a typical approach to estimation of disparity maps based on matching measures has been limited in favour of the solution in question. Both the results thus obtained and statistical analysis have confirmed legitimacy of the solution devised by the authors.