Font Size: a A A

3D Model Retrieval Based On Mesh Segmentation And Hierarchy Feature

Posted on:2014-01-20Degree:MasterType:Thesis
Country:ChinaCandidate:H F XuFull Text:PDF
GTID:2248330398458032Subject:Computer software and theory
Abstract/Summary:PDF Full Text Request
With the development of the computer software and hardware, the rapid popularization ofthe Internet and the increasing improvement of the computer graphics theory,3D mesh modelingtechnique have been widely used in many fields such as the film and television animation,simulation, biomedical and other.3D replacing2D, virtual simulation of realistic3D modelingtechnology leads people into the3D world. However, modeling highly realistic3D model is toodifficult and very laborious, if we can efficiently resue the existing3D mesh model which comesfrom Internet, we can greatly reduce the new modeling work, and how to quickly find andaccurately select the model we need has been a problem must be solved. Therefore, the researchof3D mesh modle retrieval methord has important realistic significance.The content-based retrieval is focued in3D model retrieval currently, according to3Dmodel’s materials, texture, spatial structure and other information, this methord can calculate andextract the shape feature to mark the3D model only, then calculate shape difference between thetarget model and the query model in database of3D models, and output the first N models withthe smallest difference as the result,3D model retrieval finish. Therefore, the key of3D modelretrieval is the shape feature how to be extracted.The retrieval system is analyzed and the existing featrue extraction algorithm is summarizedand the improved retrieval methord is proposed by this paper. This text mainly includes thefollowing three aspects:1. Analyzed the research background and the significance of the3D model retrievalmethord, introduced the process of3D model retrieval and the framework of system, andsummarized the key techniques of3D model retrieval. There are too many algorithms to make auniform standard to classify them, we divide them into several kinds in this paper and expatiatefor each.2. When extracted the shape feature, the existing retrieval methord only calculated thewhole information of3D model, and ignored the partial information of3D model and did notfully consider the feature points of3D model, we proposed a3D model retrieval methord basedon mesh segmentation, and applied it to the retrieval system.Firstly, we compared several signal calculation methords, obtain the flatness of meshtriangles and applied it to the improved algorithm as the height function, then pretreated themesh model, and used multi-dimension scaling(MDS) to describe pose-invariant an extract thefeature points of the model, and applied the feature points as seeds to guide mesh segmentation.When finished the segmentation, in order to avoid over-segmentation, multiple rounds dyanmicweights were uesd to merge regions from local to global to make result more reasonable. Finallywe extracted the partial information of3D model to build attribute tree, and retried3D model bycomparing the degree of tree’s matching. We analied the rationality and effectiveness of theretrieval methord by retrievaling several kinds of model, Elaborate algorithm ideas, procedures,the steps of the algorithm, data structure design. Using partial information effectively the methord can obtain more precision on the same recall and improved the retrieval speed.3. At present,majority of3D model Retrievla algorithm using only single shape feature, butsingle shape feature descripition is limited, with some certain properties of3D mesh model it cangive, and can not adapt to the all models of retrieval, there are some limitations. Then weproposed the3D model Retrieval based on hierarchy feature, according to structure the hierarchyfeature,with the feedback of users, we computed the dynamic weight which the model ismatched in the3D model database. Using the methord of user’s feedback, we adjust the weightof shape feature dynamicly, and gain different threshold.Finally, when we retrieved the3D meshmodel, we firstly compared the main shape feature with threshold, then choose an weight tocompare with the other shape feature, though we search the results by comparing with themodels in3D model database.
Keywords/Search Tags:Mesh segmentation, 3D model retrieval, Feature point, Hierarchy feature, Feedback
PDF Full Text Request
Related items