Page - 77 - in Proceedings - OAGM & ARW Joint Workshop 2016 on "Computer Vision and Robotics“

GuidedSparseCameraPoseEstimation Fabian Schenk1,LudwigMohr1,MatthiasRu¨ther1, FriedrichFraundorfer1, and HorstBischof1 Institute for ComputerGraphicsand Vision GrazUniversity ofTechnology,Austria {schenk,mohr1,ruether,fraundorfer,bischof}@icg.tu-graz.ac.at Abstract In this paper, we present an idea for a sparse approach to calculate camera poses from RGB images and laser distance measurements to perform subsequent facade reconstruction. The core idea is to guide the image recording process by choosing distinctive features with the laser range finder, e.g. building or window corners. From these distinctive features, we can establish correspondences betweenviewstocomputemetricallyaccuratecameraposes fromjusta fewprecisemeasurements. In ourexperiments,weachievereasonableresults inbuilding facadereconstructionwithonlya fraction of featurescompared to standardstructure frommotion. 1. Introduction Structure from motion (SfM) has been an active research area in computer vision for decades as it is of interest in a wide range of practical applications such as robotic navigation and augmented reality. Common SfM approaches exploit a huge number of feature correspondences and finding suitable starting views poses a challenge, which is not necessarily simplified by the abundance of features. Often, this is tackled by assuming a set of ordered images or incorporating additional measurements for camera pose initialization. In a subsequent step, all the views are merged into a common global coordinate system, where the whole scene structure in 3D is calculated and refined together with the camera poses. The optimization of the final scene structure is computationally demanding due to the largeamount of correspondences overmultiplecameraviews. Figure1: OurproposedsparseSfMapproach,wherewetakeRGBimagesandlaserdistancemeasures toestimatecameraposes and asparsepoint cloud. 77