The aim of this thesis was the development of a new laser scanning software system, together with a set of new procedures and algorithm that is suitable for the task of traffic accident surveys and the automate processing of the relevant objects and scene information. To make use of the existing advantages of mobile scanning systems, a new low-cost solution for MLS was developed in this work based on a regular static laser scanning system in combination with standard sensor systems and hardware, included in standard consumer vehicles. The available systems provide the information required for the estimation of the vehicle movements, such as the mono wheel rotational speed sensors and the yaw rate sensor. All systems interact via the CAN-Bus that guarantees all relevant information to be available within fixed deadlines and thus replaces the real-time computing system, otherwise required by the MLS. The data generated by the MLS, includes the scan device's local distance measurements and coordinates as well as the globally transformed location for each point The developed procedures make use of bolh these input streams to realize a better post-processing of the data. Regarding the automation of the point cloud processing and information extraction, multiple new algorithms have been developed and implemented in this thesis. Based on a detailed analysis of the relevant scene content, the required measurement precision and possibilities for automatic, as well as user-supported processing have been examined. These gathered preconditions were used for the evaluation of different approaches on laser scan data processing. For each of the following objects and traces, new algorithms were developed from the obtained knowledge or generated via combination of existing approaches. The detection and extraction of the street surface and its corresponding boundaries is realized by a scan-line based examination of the street surface for height-jumps and variations of the surface roughness. The detected boundaries are divided into different types (curbs, walls, changes of surface roughness, limiting road markings) and evaluated regarding their likelihood of actually being part of the road boundaries. Erroneous boundaries, resulting from occlusions in the dataset are recognized via a shadow projection approach and can be excluded from the reconstruction to significantly improve the resulting surface extraction. In this thesis, a new approach for road marking reconstruction and classification is presented, based on PCA and OBB subdivision. The developed algorithm allowed the correct classification and grouping of the road marking types (continuous and dashed lane separating markings, as well as pictographic markings, such as turning arrows). Further, satisfying results were also achieved in case of severe abrasions or soiling of the road surface. In case markings were partially occluded by foreign objects, it was possible to reconstruct the missing regions and successfully classify the fragments from conclusions based on the markings preceding and succeeding neighbours.


    Access

    Access via TIB

    Check availability in my library


    Export, share and cite



    Title :

    Automatic segmentation and reconstruction of traffic accident scenarios from mobile laser scanning data


    Contributors:

    Publication date :

    2013


    Size :

    219 Seiten, Bilder, Tabellen, Quellen


    Type of media :

    Theses


    Type of material :

    Print


    Language :

    English




    Low-cost tools for virtual reconstruction of traffic accident scenarios

    Pérez, Juan Antonio / Gonçalves, Gil Rito / Morillo Barragan, Juan Ramón et al. | Elsevier | 2024

    Free access

    Complexity of Driving Scenarios Based on Traffic Accident Data

    Dong, Xinchi / Zhang, Daowen / Mu, Yaoyao et al. | Springer Verlag | 2024


    Traffic Accident Reconstruction Technology Research

    Yang, Xiao Long ;Li, Ping ;Lv, Tao | Trans Tech Publications | 2013


    Traffic Accident Reconstruction Technology Research

    Yang, Xiao-Long / Li, Ping / Lv, Tao et al. | Tema Archive | 2013


    Analysis of Traffic Accident Reconstruction Uncertainty

    Zhang, Lei / Li, Yi-bing | ASCE | 2008