Research On Design Method Of Product Scheme Based On Model Retrieval

Structural design is an important stage in the product design process,which determines the product design effect and efficiency in the later stage.At present,enterprises have generally converted from two-dimensional design to three-dimensional.If the reuse rate of the three-dimensional model is improved,not only the rapid design of the product principle scheme to the structural scheme can be realized,the design cycle of the product can be shortened,and the intelligence design of the product structure scheme can be gradually realized.This paper studies the fast recognition method of 3D model and establishes a product design system based on model retrieval.The system supports query methods based on 2D sketches and 3D models and text keywords.The main research contents are as follows:Based on VBA's secondary development of Solidworks software,the batch output of 3D model thumbnails is realized.By extracting Zernike moments and HU moments of 3D model thumbnails and PCA feature codes as image feature descriptors and comparing them,the final selection is made.The Zernike moment of the model thumbnail is used as its feature descriptor.By manually labeling each category model in the 3D model sample library,combined with the 3D model content feature similarity and WordNet-based semantic similarity calculation rules,the label vocabulary of the model to be labeled is determined together,and the semantic verification is automatically verified by experimental verification.The algorithm has higher semantic annotation accuracy.Based on the BP neural network in machine learning,a three-dimensional model classification retrieval method is proposed.By extracting the extracted 3D model feature vector into the input of the classifier,the category number of the manually labeled partial 3D model is taken as the output.The input and output are sent to the classifier for simulation training and experimental verification.The verification results show that the 3D model to be classified can achieve more accurate classification.By automatically semantically annotating the 3D model,the 3D model can be retrieved by inputting the text keyword.In addition,the system can also input the model 3D map or 2D sketch,and realize the 3D model retrieval based on the feature similarity calculation.At the same time,due to the establishment of the model classifier,the model and the new model in the 3D model library are automatically classified,which improves the retrieval efficiency of the retrieval system.