Font Size: a A A

Research On Related Techniques Of Sketch-Based 3D Modeling

Posted on:2017-10-05Degree:DoctorType:Dissertation
Country:ChinaCandidate:K LiuFull Text:PDF
GTID:1488304841978499Subject:Computer application technology
Abstract/Summary:PDF Full Text Request
As an important method of computer-aided modeling,sketch-based 3D modeling is one of the most challenging tasks in computer graphics.The research aims to infer 3D models from the sketches in accordance with the human vision rules and object domain knowledge based on sketch recognition and understanding.Sketch-based 3D modeling is built on many existing research subjects,and has been widely used in many fields such as industry design,film animation,distance education,auxiliary medical field and so on.In this thesis,according to application requirements of sketch-based 3D modeling,we focus on several issues including diversity of sketchy drawing,variety of model type and flexibility of semantic classification,and research on several related techniques,such as sketch-based 3D face modeling,sketch-based implicit surface generation,and iterative sketch annotation.Regarding these issues,this thesis engages in the following work:(1)We proposed a sketch-based face modeling method based on pose estimation and shape parameter mapping.In this method,pose estimation is introduced to analysis face sketch,which can generate the front-view face sketch from the side-view sketch,so that users can draw the free-form sketch of the portrait from their chosen view.Then,shape parameter mapping is adopted to establish one-to-one correspondence between the sketch feature points and 3D face feature points.It allows the user to draw many different kinds of contours.Meanwhile,the displacements between the feature representatives are calculated to generate the specified 3D face models.It guarantees that the sketches' geometrical characteristics can be retained in the final 3D face model.Thus,by increasing the freedom of view and types of contour to effectively design 3D face shape,it adapts to the requirements of the diversity of sketchy drawing in sketch-based 3D modeling.(2)We proposed a sketch-based variational implicit surface generation method based on multi-view strokes and localization of variational implicit function.In this method,inflation transformation,curve spatial transformation and bend transformation are defined respectively to map three kinds of strokes(contour stroke,cross-section stroke and skeleton stroke)to 3D constraint points.It results that users can control the 3D shapes with freehand sketches,which enriches the ways of designing the implicit surface.Furthermore,the localization of variational implicit function is adopted to implement surface blending,which achieves a more effective control of the blending shape of multiple implicit surfaces.Through combining the generation of various shapes and the blending of multiple implicit surfaces,thereby the variety of the generated models from sketchy drawing with implicit surfaces is increased.(3)We propose an iterative sketch annotation method based on semi-supervised clustering and user intervention.In this method,sketch collections are recurrently annotated online in group by group form to discover the sketch categories,which removes the dependence of pre-labeled samples and pre-trained classifiers.Besides,the user can confirm the required members of each group and annotate it with any free label without any prior determinations,which makes the annotation result personal to improve the adaptation of the annotation result for different applications.Moreover,semi-supervised clustering is used to improve the clustering results,which reduces the interactive effort of iterative annotation process.Accordingly,by integrating online metric learning,semi-supervised clustering and user intervention into a whole framework,the efficiency of sketch annotation is improved while the flexibility of sketch classification is maintained.
Keywords/Search Tags:Sketch-based Interface, 3D Modeling, 3D Face Modeling, 3D Surface Generation, Variational Implicit Surface, Sketch Classification, Semi-supervised Clustering
PDF Full Text Request
Related items