Font Size: a A A

Research On 3D Indoor Scene Synthesis Technology Based On Style Consistent Retrieval

Posted on:2020-01-18Degree:MasterType:Thesis
Country:ChinaCandidate:P Y WangFull Text:PDF
GTID:2382330575452502Subject:Computer technology
Abstract/Summary:PDF Full Text Request
The emergence of the indoor home design platform simplifies the design process,reduces the operation difficulty,and synthesizes the indoor scene layout that meets the actual user needs.However,how to automatically synthesis an indoor 3D indoor scene with beautiful appearance and uniform style is still an unsolved problem.In the field of computer graphics,because there is no good 3D furniture model style analysis method,the 3D indoor scene synthesis method can only consider how to synthesize a reasonable 3D scene without considering the style consistency of 3D furniture objects.This paper proposes a 3D furniture model style analysis based on metric learning and a scene synthesis method based on style consistency retrieval.Based on this,a corresponding prototype system is designed.The main results of this paper include the following three points:(1)Propose a 3D furniture model style analysis method based on metric learning.This paper designs a quadruplet based depth metric learning network.The network consists of two feature extraction modules and a quadruplet-based metric learning module.The two feature extraction modules are 3D model mesh data shape feature extraction and 3D model.Multi-view projection feature extraction module.Based on the existing network frameworks with various style characteristics,the style consistency characteristics of the 3D model can be analyzed.(2)Propose a 3D indoor scene synthesis method based on co-retrieval.This paper proposes a 3D scene synthesis method based on style consistency retrieval.The furniture style collaborative search can be carried out by combining the model style features of the furniture model with the scene layout information,and the retrieved furniture model can be placed in a reasonable position to ensure the position and layout of the final furniture model.The information is consistent.This paper uses the constraint method of furniture model position adjustment.In the scene synthesis process,the method can ensure the style consistency of the furniture model of the synthetic scene,the position and size of the scene furniture model should be consistent with the scene layout information provided by the user.(3)3D indoor scene synthesis prototype system.In this paper,the 3D furniture model style analysis method is integrated with the 3D indoor scene synthesis method based on style consistency,and the prototype system of 3D indoor scene synthesis is designed and implemented.The user needs to input a 3D indoor scene and a target style furniture object,and the system automatically realizes the synthesis of the furniture model style and the final scene.This paper first introduces the relevant modules of the system,and shows the use process and implementation effect of the system through examples.The system can be used as a tool for furniture style analysis,and it can also help users quickly synthesize 3D indoor scenes with consistent style.
Keywords/Search Tags:3D Shape Analysis, Scene Synthesis, Style Analysis, Deep Learning, Style Consistency
PDF Full Text Request
Related items