Research On 3D Model Retrieval Method For Rapid Design Of Agricultural Machinery

With the development of digital technology and computer technology,the model of product design has changed.Traditionally,what is mainly considered is how to build a three-dimensional model in designing products.Now,the main consideration in designing products is how to find the three-dimensional model from the existing database and reuse it for increasing the efficiency of products design.Therefore,three-dimensional model retrieval technology has become a hot topic in various fields.The three-dimensional model retrieval technology is introduced into the research and development of agricultural machinery,and a technology system for rapid design of agricultural machinery is established based on modular design and other technologies.This paper discusses three key technologies and focuses on the 3D model retrieval technology as follows.(1)As the development of China's agricultural machinery is later than that of Western countries,the overall R&D capacity is still relatively lower.Particular in hilly and mountainous areas,agricultural machinery is inevitably affected by terrain and other factors,so the design process encounters more difficulties.For the sake of solution to this problem,an overall technical framework for the rapid design of agricultural machinery has been established.The components of key technology in the technical framework are elaborated,and the 3D model retrieval technology is deeply studied.The traditional 3D model retrieval method describes a single feature of model shape that only characterizes a part of the model,so it always fails to find out the right model accurately.To accurately describe the wanted model and quickly search them out,this paper focuses on the feature fusion 3D model retrieval method.(2)Through the 3D model retrieval method based on a projected view,the retrieval is inefficient because it takes a long time in massive calculation in the process of voxeliztion,which greatly decreases the efficiency of the retrieval.Therefore,a model voxelization method based on extended triangle boundary is proposed.Through the outward extension of the triangle boundary,the voxelization of points,lines and surfaces is taken place by the voxeliation of surfaces,which can reduce the calculation of the distance between spatial points and lines.The comparative experiment shows that this method greatly shortens the time of the model voxelization.(3)The traditional image feature extraction algorithm based on Fourier transform uses only the amplitude information of the Fourier transform coefficients but not the phase information,thus a boatload of shape features are missing.Therefore,the feature extraction algorithm based on Fourier transform using the phase retention is applied to extract the image features,which simultaneously retains both the phase information and the amplitude information.The experimental results show that,compared with the traditional feature extraction algorithm based on Fourier transform,this algorithm got more ideal retrieval results.(4)To cover the shortcomings of single features descriptor in the description of model shape,features of Fourier transform using phase retention and the Zernike moment features are integrated by the information entropy,which creates complementary advantages.The experimental result shows that the method of feature fusion works better in retrieval than the single feature extraction method.(5)As has been complemented above,a 3D model retrieval interface for rapid design of agricultural machinery is designed.Three retrieval methods,including Fourier transform feature,Zernike moment feature and feature fusion,are integrated and implemented.Applying this retrieval interface,it is achievable to preview and download the retrieval results and it is also feasible to conduct the second retrieval.