The defense of his dissertation in mid-November at the University of Saint-Étienne marked the successful completion of Jean-Jacques Ponciano’s doctoral thesis. He was advised by Prof. Dr. Alain Trémeau of the University of Saint-Étienne and Prof. Dr. Frank Boochs of Mainz University of Applied Sciences. Congratulations, Jean-Jacques!
In his dissertation, titled “Object Detection in Unstructured 3D Data Sets Using Explicit Semantics,” Ponciano first examined existing methods that automatically detect objects contained in 3D point clouds through appropriate processing. His conclusion: The best method depends on the specific context, the type and quality of the data to be processed, and the objects to be detected. However, the need to adapt a method to a specific use case limits its transferability to other fields.
In his dissertation, Jean-Jacques Ponciano overcomes this limitation with a knowledge-based approach to object recognition that can be applied independently of the specific application area. His architecture is based on semantic technologies that enable a knowledge management module to guide the object recognition process through a step-by-step procedure for selecting, parameterizing, and executing algorithms. The recognition process is carried out using an artificial intelligence approach that employs explicit knowledge to flexibly integrate context into the solution. In addition to this adaptability, the approach is also capable of analyzing and understanding a scene, the objects it contains, and the specific characteristics of the data to be processed. This capability is achieved through a self-learning process that can define and validate hypotheses about the context, thereby expanding the knowledge base and improving the object recognition process.
The efficiency of this method is demonstrated in four applications from different fields: In the context of BIM, for example, the interior of a building is broken down into its spatial structure. Representing the field of archaeology are the ancient ruins of a building complex in Ephesus, where a watermill present on-site is automatically detected through data processing and inference. In the context of mobile mapping, a section of the city of Freiburg was segmented. The final application is an indoor project acquired from Microsoft Kinect. It is used for robotic purposes.
The research results from Ponciano’s dissertation open up a range of possibilities that will also be applied and further developed in additional research projects at i3mainz.
