Sarah Schütz and Lukas Haas from the Department of Geoinformatics and Surveying were awarded the 2023 VDI Rheingau Regional Chapter Incentive Prize on May 4, 2023, for their outstanding theses.
In his master’s thesis, titled “Development of Algorithms for Automated Segmentation and Approximation of Point Clouds,” Lukas Haas set out to automatically identify individual objects from the vast amount of data in an unstructured point cloud, delineate them from one another, and model the individual objects in three dimensions. Until now, the point cloud has been laboriously trimmed by hand to fit each individual object.
His approach structures the point cloud hierarchically in an octree, recursively subdividing the volume elements (voxels) until the termination criterion is met. The points contained within the volume element are approximated using a second-degree surface (quadric). The resulting approximation accuracy is used to further prioritize the voxels. If points from neighboring voxels match the approximated quadric, they are added to the prioritized voxel, so that the region of the prioritized voxel gradually expands. The list of prioritized voxels is processed iteratively until the point cloud is divided into homogeneous segments, each of which is mathematically modeled by the quadric parameters.
This approach was fully implemented in the Python programming language and tested using simulated and measured point clouds with various object groups and arrangements of differing levels of detail. The quality of the results is very good; however, as is also the case with manual evaluation, points cannot be correctly assigned at the boundaries between two objects. Nevertheless, the algorithm is ideally suited as a tool to help segment point clouds of existing objects and approximate geometric shapes.
In her bachelor’s thesis on the topic of “Automated Determination and Provision of the Current Traffic Situation through Machine Image Recognition,” Sarah Schütz aimed to develop a workflow that determines traffic volume. This enables, for example, the integration of traffic volume data into existing systems to implement more efficient route planning for pedestrians, cyclists, or drivers. With the information obtained about current traffic conditions, road users can thus avoid road closures, heavily congested segments of the road network, or areas with high traffic volumes. The prototype approach determines traffic flow using the Internet of Things (IoT) in conjunction with image recognition and freely available webcams. For Mainz, the capital of Rhineland-Palatinate, the installed webcams were able to provide the necessary traffic data for an initial test. “In her bachelor’s thesis, Sarah Schütz was able to implement a prototype that is not only effective but, above all, sustainable and openly accessible through the use of open-source libraries and services,” said her advisor, Pascal Neis.
We, as a department, extend our warmest congratulations to the two honorees and wish them continued success!
