Research On Automatic Pipe Network Modeling Based On 3D Point Cloud Data

As the important urban infrastructure, pipeline bears people‘s daily life by material transformation and waste excretion. Construction and improvement of pipe network provide strong support for our decision in urban planning, construction and management. The traditional two-dimensional spatial network cannot satisfy our needs for city planning and construction management and factory management in the modern urban area because of its poor spatial performance, lack of spatial query analysis capabilities and ineffectively complex network model presentation. Thus, it is urgent to rebuild the real three-dimensional model of the pipe. The 3D laser scanning can acquire pipeline data in all-weather and high-precision means relative to traditional point measurement.Developed rapidly in recent years, 3D laser scanning has become an indispensable means to obtain real three-dimensional coordinate data. Laser radar scanning system is capable of obtaining high precision massive point clouds in a short time. The amount of data arranges from several G to several hundred G, even to the level of TB size. So it is critical for laser radar technology development and applications in terms of how to proceed the scanned data precisely. However, the disorder of the point cloud data increases the processing difficulty. We study how to reconstruct the three-dimensional model of the pipe network from the perspective of the point cloud data for factories and urban modernization construction applications.Currently, the network modeling technique is mainly based on manual or semi-manual modeling standard connectors, and the automatic three-dimensional pipe network modeling could be insufficient in modeling complex, inefficient, inadequate and poor quality of the model. To solve the above problem, we propose an automatic process of the pipe network reconstruction based on the point cloud data. In this process, we first estimate the parameter of a single pipe by dividing complex data of the pipe network, and then reconstruct the comprehensive pipeline with the modeling single pipe. In this paper, the process of the algorithm consists of four steps: pipeline point cloud data division, single pipe parameterization, analysis topology between the pipe networks, and reconstruction of the pipeline model parameters.First, according to the similar characteristics of pipe surface points, this paper presents a method of splitting into a single pipeline and removing data other than pipeline by initial spatial clustering based on the local feature point cloud data space constraints. Complex pipe network is simplified into a single pipeline to model, effectively reducing the complexity of modeling. The process of parameterizing a single pipeline consists of three steps, feature point of pipeline centerline extraction, pipeline parameter fitting and mutation detection. In this paper, L1-Mean algorithm is used to extract the pipeline skeleton. Then, we slice the data to fit pipeline parameters along skeleton direction, and finally do the mutation detection for the pipeline with mutations based on geometric projection.After extracting the parameter of the single pipeline, we need to make the necessary pipeline topology analysis, including the same pipeline detection and cross detection.In the process of reconstruction of the parameter model, an algorithm for the generation of complex pipeline model based on sweeping, lofting and Boolean operations was proposed. According to parametric curves of the sectional shape, we determine the normal of the cross-section using graphics transformation. Then we can realize the model construction based on sweeping along the path. Loft constructs the pipeline surface with triangle plane composed of vertices of the loft sections, using the method of approximation to fit the overall shape. the construction of three-dimensional model for the complex pipe network which cross through each other by using Boolean operations union, intersection and subtraction operation.In this paper, the algorithms and techniques constitute a complete process.that can automatically use a large amount of point cloud data of pipeline system to generate three-dimensional model. By instance, our result show that the method has some robustness, and it can construct complex 3D pipeline model accurately and efficiently.