3D GIS Model Based On BIM Data Source And The Research Of Its Application

There is an increasing attention on the integration of Building Information Model(BIM)and GIS.The data models derived from the integration of BIM and GIS have been widely used in many application domains,such as the management of infrastructure,disaster evaluation,indoor navigation and architecture engineering construction(AEC).BIM provides timely,accurate and enough information for the construction of a project's life cycle at different stages and for different parties,it mainly focuses on the spatio-temporal planning and management.Whereas,GIS focuses on the description of the already existed spatio-temporal objects.Therefore,the recognitions of BIM and GIS about the spatio-temporal objects are different.The cognitive differences lead to a big semantic gap and take great challenge to the integration of BIM and GIS.In order to make some progress in this field,this paper make a study about the integration of Industry Foundation Classes(IFC)and City GML which is the most representative semantic model in BIM and GIS world,respectively.In this paper,semantic integration,the method of coordinate transformation(local coordinate and the global geographic coordinate system),the method of geometric transformation and the corresponding application have been studied.Also,the studies about the semantic annotation method based on the 3D model classification,3D spatial partition algorithm have been made.The main results are showed as follows:(1)Proposed a semantic matching and filtering method based on texture mining techniqueIn this paper,a texture mining technique based semantic matching method was used to match the semantic entities defined in the IFC 2×3 schema with those of defined in City GML 2.0 schema.Correct semantic matches were provided by this method and mismatches were filtered out,the level of the integration automation between the two schemas was thus improved.In this method,a new word vector IIgenerator based on the word-hashing was proposed.The processes of stop words remove and word stemming in the traditional word vector generator were omitted in the new one.With the same semantic similarity measurement,the precision and the recall rate of the semantic matching provided by new word vector generator was higher than that of the traditional one,respectively.Taking both the feature of semantic entity name and definition into account,the highest semantic matching precision and recall rate derived from the new word vector generator with the cosine similarity measurement was 100% when ranking threshold value was lower than 60.The semantic matching precision decreased with the increase of ranking threshold value,while the recall rate increased with it.The highest recall rate of 0.821 was also achieved by the new word vector generator at the case of ranking threshold value was200.The experiment results showed that the proposed method outperformed the traditional one.(2)Transformation of coordinate system and geometric structureThe coordinate system and the geometric structure of IFC and City GML were analyzed.The performances Taylor series expansion-based transformation method and the Roderick matrix-based transformation method were compared and introduced into the integration process of IFC and City GML.For the case of rotation with large angle,the performance of the Roderick matrix-based transformation method was significantly better than that of Taylor series expansion-based method.In terms of the geometric structure,a geometric transformation method was proposed to parse and transform the swept solid of IFC to the boundary representation model(B-rep)of City GML.According to the schema of IFC and City GML,an intermediate ontology was proposed and the integration of IFC into City GML was completed.(3)semantic annotation of 3D data model and 3D spatial partition algorithmDue to the semantic missing of most 3D mesh models,the geometrical analysis methods of computer graphics were introduced into the fields of semantic annotation and 3D spatial partition.The shape feature of the 3D mesh model was extracted by the ray-based feature extraction method.Then,the semantic labels of 3D mesh models were assigned by classifiers of support vector machine(SVM)and extreme machinelearning(ELM).In order to avoid the inefficiencies of 3D model indexing and intersection,an ELM was applied to divide the 3D mesh model into several subparts.In the process of 3D mesh model segmentation,the shape descriptors of geodesic distance and shape diameter function were constructed to feed the ELM.The experimental results showed that the efficiency of the 3D mesh model intersection with the divided model was greatly improved.(4)Path planning algorithm based on the 3D data modelIn the first place,a separating axis theorem-based 3D model voxelization method was used to divide the 3D space into a series of regular grids.Then a kind of Hash Map-based linear octree was built and used to construct octree map for indoor path planning.In addition,the A star(A*)algorithm which was often used to search the shortest path in 2D space was extended to 3D space in this paper.In the indoor path planning process,a 26-neighborhoods model was used as the searching space of the A* algorithm.The experimental results demonstrated that the algorithm can find the shortest path and perfectly keep away from the spatial obstacles.In this paper,the semantic matching and filtering out methods,coordinate and geometry transformation methods,algorithms for semantic annotation of 3D model,and model segmentation were studied.These works made a firm foundation for the further integration of BIM and GIS.The path planning algorithm based on the 3D model demonstrated a good application.