POINT CLOUD SEGMENTATION USING IMAGE PROCESSING TECHNIQUES FOR STRUCTURAL ANALYSIS

Modern surveying techniques, with the combined use of Unmanned Aerial Vehicles (UAV) with low-cost photographic sensors, and photogrammetric techniques, allows obtaining a precise virtual reconstruction of environment and object with centimetre accuracy. Recently, the diffusion of UAV allows the survey of extensive areas significantly reducing survey time and costs. The raw output obtainable from such survey operations consists of a three-dimensional point cloud. Numerous applications in architecture, monitoring and surveying and structural analysis require objects identification in the 3d scene to classify different element in the acquired scene and extract relevant information. Point cloud analysis, and in particular segmentation and classification techniques, are actually used to identify objects within the scenes, assign to a specific class and use them for subsequent studies. These techniques represent an open research theme and the key to add value to the entire process. Actual methodologies are based on 3d spatial analysis on the point cloud. In this paper, starting from photogrammetric reconstruction, a methodology for segmentation and classification of point cloud based on image analysis is presented. The object identification on the image’s dataset is performed using a Neural Network and subsequently the identified object on dataset are transfer into the 3d environment. This classification is performed to segment structural parts of bridges and viaduct, acquire geometric information, and perform a structural analysis to preserve relevant and ancient structure. A case study for the segmentation of the point cloud acquired with an aerial survey of a Viaduct is presented. The performed segmentation allows obtaining structural elements of different type of viaduct and bridges, is propaedeutic to verify the health of the structure and schedule maintenance intervention. The methodology can be applied to different type of bridges, from reinforced concrete to ancient masonry to preserve the state of conservation.