Line Segments Matching Algorithm for BIM Applications

A geometric image matching approach, based on line segments, is proposed to compare simulated and real images. Line segments are first clustered based on their proximity and absolute angle. The centroids of the clusters are then used as input for the image matching. As an initial step, a set of features is extracted from each line segment and used to find the best match. This matching is then exploited to identify the local arrangement of the other lines in the simulated and real image. Since the descriptors rely only on the local geometry of the line segments, the method performs robustly against the diverse nature of the images involved. Many simulations and comparisons with the state of the art are provided to evaluate the characteristics of the proposed method, which is equivalent in performance when applied to images provided by the same source and very accurate with images from different sources. This features is desirable, for example, when it is required to compare images obtained from rendering a 3D BIM model with those obtained from a camera.