A piecewise affine warping algorithm for image rectification in an industrial application for ship hull repair

Ship hull maintenance is a periodical task. It requires working conditions extremely hard for operators. The authors have developed a robotized system that makes it possible to carry out these tasks remotely and semi-automatically, by means of an automated visual inspection system. Due to the hard work environment conditions the vision system needs to be placed far from the front and central position opposite the surface to be inspected. As a consequence, the produced images show strong perspective distortions, which need to be corrected. In this paper we present an image correction algorithm, based on affine warping transformation techniques, which allows us to correct distortions in images of strongly warped surfaces. In addition, this algorithm provides a final image equivalent to a front image consisting of the image fragments that the camera would obtain should it be placed opposite each of the points of the inspected surface from which the image is taken. The algorithm produces high precision corrected images which make it possible to locate any point in that surface with a maximum error of 3 cm over a 200×160cm2 surface (1.17%). We have used this algorithm to correct images of ship surfaces with different profiles and curvatures, where traditional techniques based on calibration grids are not applicable due to the unavailability of a grid pattern for each new surface to be inspected.