Studies On3D Solid Reconstruction From2D Engineering Drawings With Sectional Views

3D solid reconstruction from2D engineering drawings is one of the key technolo-gies in informatization of industry and manufacturing. It is also an important researchtopic in the field of computer aided design (CAD). This dissertation discusses several keytechnologies in reconstructing3D models from sectional views.The main work is summarized as follows.1An algorithm is proposed to handle sectional views by improving the traditionalvolume based method. First, we present a two-stage loop searching algorithm to extrac-t desired loops from both sectional views and non-sectional views. Then, sub-objectsare identified by the hint-based identification algorithm with an intuitive loop-matchingcriterion. After that, a model-directed algorithm is proposed to guide the generation ofsub-objects which are assembled together to form the final objects. Multiple sectionalviews are supported in the algorithm. Moreover, the domain of objects is extended toinclined quadric surfaces and intersecting quadric surfaces with higher order curves.2A novel method is proposed to identify, construct and validate sub-objects fromsectional views. First, sub-objects are classified as explicit sub-objects (EPFs) and implic-it sub-objects (IPFs), which are then identified in an order of priority using heuristic hints.We show that the problem of constructing EPFs can be formulated as a0-1integer linearprogram (ILP), and the IPFs are generated based on the understanding of semantic infor-mation of omitted projections in sectional views. Then, the Sub-objects-Relation Graph(SRG) is introduced as a multi-connected-subgraph representation for describing the rela-tions between loops and sub-objects. According to the SRG, a reasoning technique basedon confidence is implemented to interactively validate sub-objects. This method can re-cover sub-objects without complete projections, and the level of understanding sectionalviews is improved. Full sections, half sections, partial sections, ofset sections as well asrevolved sections can be handled by the method.3A new algorithm is presented to understand the semantic information of shaftsfrom their2D drawings, and then reconstruct the3D models. The2D representationsof shafts are diverse. By analyzing the characteristics of2D drawings of shafts, we findthat the problem of identifying elemental parts can be transformed into a classification problem. The conditional random fields (CRFs) model is a classification technique whichcan automatically integrate various features, rather than manually organizing of heuristicrules. We first use a CRFs model to identify elemental parts with semantic information.The3D elemental parts are then constructed by a proposed parameters template method.Single view, coincide sectional view and removed sectional view can be handled by thealgorithm. Compared with the existing3D reconstruction methods, this approach canobtain both geometrical information and semantic names of each part of shafts from2Ddrawings.