ABSTRACT
visual inspection of highways bridges, viaducts and tunnels has nowadays a crucial role in infrastructures life-cycle management. In this paper two visual inspection numerical support methods are discussed. Both methods aim to reduce direct and indirect costs of visual inspection and to automate the inspection procedure. This is possible thanks laser scanner images of the tunnel, which are quickly acquired with remarkable decrease of money investment and with no need for in loco presence of engineers and then used as input for the calculations. The former method is based on Dynamic Modal Decomposition and it uses laser scanner acquisition to predict damage evolution. The latter one is based on a convolutional neural network and it aims to evaluate damage gravity according to a predefined severity scale. Both methods have been tested on Italian highways tunnels images.
