A Feature-based Matching Strategy for Automated 3d Model Reconstruction in Multi-image Close-range Photogrammetry

Feature-based matching is commonly employed for object surface reconstruction in topographic and stereo close-range photogrammetry, but rarely in conjunction with convergent photogrammetric networks. This paper describes a new feature-based matching approach to automated 3D object reconstruction from highly convergent, multi-image networks. Geometric diversity and redundancy in the network design are a distinct advantage with the approach, which commences with automatic exterior orientation. The FAST interest operator is then applied to Wallis-filtered images to extract interest points, and matching of interest points is carried out to yield a dense 3D point cloud. The surface point data is subsequently converted to a triangulated mesh via a Poisson Surface Reconstruction technique, and textured. Three core components of the proposed object reconstruction approach, namely the FAST interest operator, Wallis filtering and Poisson Surface Reconstruction are first described. The results of an experimental test of the approach are then presented, the measurement being an engineering application in which surface modeling is performed to support deformation analysis. The results obtained highlight the practicability, robustness and accuracy of the reported automated multi-image object surface measurement approach incorporating feature-based matching.