Maintenance management in the road sector currently uses information and data from the condition recording and assessment (ZEB) of the road surface. Together with further information on the road network (inventory, measures, costs), these represent one of the foundations of modern BIM architectures in road management. Current methods are based on sensor measurements of evenness and skid resistance as well as on systematized, visual inspection of images of the road surface. The acquisition and especially the evaluation of the data is time and manpower intensive. Partial or interim results can only be made available at considerable expense due to the nature of the process. The use of a Ground Penetrating Radar (GPR, also Georadar) allows the non-destructive recording of the road structure as well as the detection of anomalies, but is currently only used in special cases. The main hurdle to the extensive application of this method is the manual effort required for the acquisition and, in particular, the evaluation of the collected data. In the BMDV-funded project RADSPOT, a prototype measurement vehicle was created that can collect environmental data by means of LIDAR sensors and camera recordings and, as a central innovation, enables the fully automated collection and processing of georeferenced GPR data. Also demonstrated was the automated (live) data flow from mobile vehicles to a central cloud instance. The application of the created system was successfully demonstrated for selected road clusters in the Kaiserslautern area.

In the research part of the project, ZEB surface images as well as GPR data are subsequently analyzed using developed state-of-the-art AI and ML algorithms of pattern recognition. All data will be stored locally and transmitted to the Digital Twin via internet in encrypted form. Any personal information (faces, license plates, etc.) present in the dataset will be anonymized using current techniques. The project goal is to create a Digital Twin platform that automate storage, processing, and interpretation to the large amount of collected data. This enables the use of the generated information by users by linking ZEB and GPR data through an integrative data management and interpretation system. The datasets are managed using georeferencing and sensor timestamps, aggregated both spatially and temporally. Moreover, the data is based on a topological network model to which topographical information is assigned for cartographic visualization. In addition, the data collected in the digital twin are enriched with ontologies and knowledge graphs that semantically describe the measurement data collected by the test vehicle. To evaluate the developed AI-based algorithm, it must be assessed whether the targeted data volume allows sufficient recognition accuracy using AI. The problem can be addressed by collecting additional data (GPR), whereby the implemented automation massively reduces the effort for this. Further issues arise from the consideration of whether the technically possible location accuracy of all data is sufficient for synchronize and aggregate ZEB and GPR data. However, due to the novelty of this aggregation process, this question is more of a research topic than a technical hurdle.

• Automated processing and creation of a reference data, labeling of sufficient object features and AI-based interpretation of this big dataset for the entire Kaiserslautern urban area, including all other data sources involved in the project.
• Development of a service for automated processing of ZEB data.
• Multi-sensor data integration and synchronization between linearly referenced data and geo-referenced data in one Digital Twin, as well as further development of the data representation to 2.5D / 3D- representation, in order to increase the precision of detected features.
• Automated evaluation of the surface data derived from the ZEB survey using AI and ML based methods, merging the results with the detections of the georadar data in the Digital Twin to enable an integrative interpretation, evaluated according to professional parameters.

• Digital Twin
• ML-based Feature Detection
• Ground Penetrating Radar (GPR)
• ZEB Data Analytics
• Cloud Services

Projektträger: mFUND

 

Jan 2023 - Dec 2024

• Technische Universität Kaiserslautern
• HELLER Ingenieurgesellschaft mbH
• Geophysik GGD mbH
• Stadt Kaiserslautern

Prof. Dr.-Ing. Naim Bajcinca
Gottlieb-Daimler-Str. 42
67663, Kaiserslautern
+49 (0)631/205-3230
naim.bajcinca(at)mv.uni-kl.de