<> "The repository administrator has not yet configured an RDF license."^^ . <> . . "Model-Based Multiple 3D Object Recognition in Range Data"^^ . "Vision guided systems are relevant for many industrial application areas, including manufacturing, medicine, service robots etc. A task common to these applications consists of detecting and localizing known objects in cluttered scenes. This amounts to solve the \"chicken and egg\" problem consisting of data assignment and parameter estimation, that is to localize an object and to determine its pose. In this work, we consider computer vision techniques for the special scenario of industrial bin-picking applications where the goal is to accurately estimate the positions of multiple instances of arbitrary, known objects that are randomly assembled in a bin. Although a-priori knowledge of the objects simplifies the problem, model symmetries, mutual occlusion as well as noise, unstructured measurements and run-time constraints render the problem far from being trivial. A common strategy to cope with this problem is to apply a two-step approach that consists of rough initialization estimation for each objects' position followed by subsequent refinement steps. Established initialization procedures only take into account single objects, however. Hence, they cannot resolve contextual constraints caused by multiple object instances and thus yield poor estimates of the objects' pose in many settings. Inaccurate initial configurations, on the other hand, cause state-of-the-art refinement algorithms to be unable to identify the objects' pose, such that the entire two-step approach is likely to fail. In this thesis, we propose a novel approach for obtaining initial estimates of all object positions jointly. Additionally, we investigate a new local, individual refinement procedure that copes with the shortcomings of state-of-the-art approaches while yielding fast and accurate registration results as well as a large region of attraction. Both stages are designed using advanced numerical techniques such as large-scale convex programming and geometric optimization on the curved space of Euclidean transformations, respectively. They complement each other in that conflicting interpretations are resolved through non-local convex processing, followed by accurate non-convex local optimization based on sufficiently good initializations. Exhaustive numerical evaluation on artificial and real-world measurements experimentally confirms the proposed two-step approach and demonstrates the robustness to noise, unstructured measurements and occlusions as well as showing the potential to meet run-time constraints of real-world industrial applications."^^ . "2010" . . . . . . . . "Dirk"^^ . "Breitenreicher"^^ . "Dirk Breitenreicher"^^ . . . . . . "Model-Based Multiple 3D Object Recognition in Range Data (PDF)"^^ . . . "thesis_breitenreicher.pdf"^^ . . . "Model-Based Multiple 3D Object Recognition in Range Data (Other)"^^ . . . . . . "indexcodes.txt"^^ . . . "Model-Based Multiple 3D Object Recognition in Range Data (Other)"^^ . . . . . . "lightbox.jpg"^^ . . . "Model-Based Multiple 3D Object Recognition in Range Data (Other)"^^ . . . . . . "preview.jpg"^^ . . . "Model-Based Multiple 3D Object Recognition in Range Data (Other)"^^ . . . . . . "medium.jpg"^^ . . . "Model-Based Multiple 3D Object Recognition in Range Data (Other)"^^ . . . . . . "small.jpg"^^ . . "HTML Summary of #10582 \n\nModel-Based Multiple 3D Object Recognition in Range Data\n\n" . "text/html" . . . "004 Informatik"@de . "004 Data processing Computer science"@en . .