Sensing of complex buildings and reconstruction into photo-realistic 3D models

F.J. Heredia Soriano, Technische Universiteit Eindhoven (TUE). Stan Ackermans Instituut. Software Technology (ST)

    Onderzoeksoutput: ScriptiePd Eng Thesis

    356 Downloads (Pure)

    Samenvatting

    The 3D reconstruction of indoor and outdoor environments has received an interest only recently, as companies began to recognize that using reconstructed models is a way to generate revenue through location-based services and advertisements. A great amount of research has been done in the field of 3D reconstruction, and one of the latest and most promising applications is Kinect Fusion, which was developed by Microsoft Research. Its strong points are the real-time intuitive 3D reconstruction, interactive frame rate, the level of detail in the models, and the availability of the hardware and software for researchers and enthusiasts. A representative effort towards 3D reconstruction is the Point Cloud Library (PCL). PCL is a large scale, open project for 2D/3D image and point cloud processing. On December 2011, PCL made available an implementation of Kinect Fusion, namely KinFu. KinFu emulates the functionality provided in Kinect Fusion. However, both implementations have two major limitations: 1. The real-time reconstruction takes place only within a cube with a size of 3 meters per axis. The cube's position is fixed at the start of execution, and any object outside of this cube is not integrated into the reconstructed model. Therefore the volume that can be scanned is always limited by the size of the cube. It is possible to manually align many small-size cubes into a single large model, however this is a time-consuming and difficult task, especially when the meshes have complex topologies and high polygon count, as is the case with the meshes obtained from KinFu. 2. The output mesh does not have any color textures. There are some at-tempts to add color in the output point cloud; however, the resulting effect is not photo-realistic. Applying photo-realistic textures to a model can enhance the user experience, even when the model has a simple topology. The main goal of this project is to design and implement a system that captures large indoor environments and generates 3D photo-realistic large indoor models in real time. This report describes an extended version of the KinFu system. The extensions overcome the scalability and texture reconstruction limitations using commodity hardware and open-source software. The complete hardware setup used in this project is worth €2,000, which is comparable to the cost of a single professional laser scanner. The software is released under BSD license, which makes it completely free to use and commercialize. The system has been integrated into the open-source PCL project. The immediate benefits are three-fold: the system becomes a potential industry standard, it is maintained and extended by many developers around the world with no addition-al cost to the VCA group, and it can reduce the application development time by reusing numerous state-of-the-art algorithms.
    Originele taal-2Engels
    Toekennende instantie
    Begeleider(s)/adviseur
    • Jalba, Andrei C., Begeleider
    • Bondarau, Egor, Externe begeleider
    Datum van toekenning4 okt 2012
    Plaats van publicatieEindhoven
    Uitgever
    Gedrukte ISBN's978-90-444-1162-1
    StatusGepubliceerd - 2012

    Bibliografische nota

    Eindverslag

    Vingerafdruk Duik in de onderzoeksthema's van 'Sensing of complex buildings and reconstruction into photo-realistic 3D models'. Samen vormen ze een unieke vingerafdruk.

  • Citeer dit

    Heredia Soriano, F. J., & Technische Universiteit Eindhoven (TUE). Stan Ackermans Instituut. Software Technology (ST) (2012). Sensing of complex buildings and reconstruction into photo-realistic 3D models. Technische Universiteit Eindhoven.