Research On Algorithm Optimization Of3d Object Retrieval Under MapReduce

Recent years, due to the massive incremental of3d model, high reliability, highaccuracy and high availability of3d model retrieval technology become an urgent need tothe reuse of3d model.3D model retrieval technology has become a hot topic in computergraphics research field. But in the existing storage capacity and computing, how to storagesuch large files on a regular server and achieve the model transformation process, featureextraction, retrieval calculations are all serious problems. In addition, due to thedevelopment of network technology, the design and sharing of3d models require renderingand processing the models on the browser. Thus problems of model files transmission onthe network and3d model rendering issues on the client appear.Firstly, the3d model retrieval technology and systems are researched, and the regularexpressions to parse the3d model documents are designed, what’s more a small Hadoopcluster is build. In the article the3d models are stored in the HDFS, taking advantage ofstorage capacity, reliable, efficient, scalable and other advantages of the Hadoop clusters todeal with the storage pressure of the rapid growth of the model files. In this thesis,3dmodel standardization, processing, feature extraction and other operations or algorithmsare designed and implemented under the MapReduce programming environment. Usingcluster’s computing resources, saving the time of3d model preprocessing and featureextraction. Without the limitation of computational resources, the accuracy of3d modelretrieval is improved. Furthermore, in order to shorten the response time and the renderingissues on the client, a3d model data compression and decompression method based on B/Sstructure is designed and achieved. The organizational structure of vertices and faces sheetis optimized, and the data of model can be extracted quickly on the client, which can bebrowsed using WebGL. Finally, a3d model retrieval system on Hadoop Big Data platformis achieved based on these three aspects above.In this paper, the3d model retrieval and the emerging big data processing method arecombined. Data compression techniques and WebGL technologies are used to improvethe performance of the3d model retrieval system. This paper promotes the development of3d model retrieval technology in some degree.