Research And Implementation Of 3D Model Retrieval Based On Deep CoNet

At present,with the continuous development of three-dimensional modeling,scanning and computer vision,the research and application of related technologies such as driverless,three-dimensional scene roaming and smart city construction have received widespread attention.As one of the most challenging tasks in scene understanding,3D model recognition has been the focus and research hotspot in the field of computer vision and computer graphics.The key to identification is how to construct an effective feature representation.Especially for complex data types such as 3D models,the original representation is unstructured and high-dimensional.How to make the computer automatically and effectively capture its essential features has always been the hotspot and difficulty of the industry.As a feature self-learning technology,deep learning enables the machine to automatically learn the multi-layer abstraction and representation of objective objects,thus understanding the intrinsic meaning of complex objects,and completing the extraction of essential features,which brings a new research direction for the retrieval of 3D models..Therefore,for the three-dimensional model retrieval problem,this paper studies and implements a three-dimensional model multi-modal retrieval method based on deep learning.A three-dimensional model multi-view classification algorithm based on deep integration learning and metric learning is proposed.By introducing metric learning into the view classification,the difficult-to-segregate samples in the view are effectively processed to improve the recognition ability of the 3D model,and the multi-view integration is ensured in the decision-making layer to ensure the three-dimensional.The rationality of model feature descriptor extraction is proposed.The multi-modal data representation and similarity evaluation algorithm based on deep learning is proposed to realize the effective retrieval of 3D models by different modal data.The 3D model multi-modal retrieval based on Deep CoNet is proposed.The new method can realize multi-modal retrieval of 3D models according to different input types such as texts,sketches and examples,improve the accuracy and validity of 3D model retrieval,andsatisfy the diversified retrieval of 3D models by users with different scenarios and different needs.demand.Based on the above research,a three-dimensional model multi-modal retrieval system based on Deep CoNet is designed and implemented under Windows.The system relies on the existing research content as the theory,and uses the 3D model standard data set as the training and test sample to realize the multi-modal retrieval of the 3D model.The data online training module is designed to prevent system changes due to database changes.A series of redundant operations.Finally,the accuracy of the algorithm and the effectiveness of the system are comprehensively evaluated in a qualitative and quantitative manner.The experimental results show that the proposed algorithm achieves excellent results in the rigid 3D model datasets ModelNet10 and ModelNet40,and the non-rigid 3D model datasets SHREC10,SHREC11,and SHREC15.The classification accuracy is high and the stability is strong.The system implemented in this paper supports the multi-modal 3D model retrieval requirements based on text,sketch,instance,etc.,which is flexible,convenient and easy to use.