Research on 3D Model Retrieval Technology Based on Audio Recognition

At present, the number of 3D models in machinery industry is increasing, and the commonly used keyword retrieval methods need a lot of manual labeling work, and many complex 3D models are difficult to describe accurately in words. This paper proposes a 3D model retrieval algorithm based on audio recognition, which can quickly and effectively find 3D models in massive models. This paper takes the 3D model of STL file format as the research object, and its main work is as follows: Firstly, a new rotation normalization algorithm for 3D models is proposed. Then a point density control scheme of STL file based on unit volume is proposed. After that, an audio fingerprint extraction algorithm for 3D model is proposed. Finally,the original speed matching method and the neighborhood matching method are proposed to calculate the similarity between models.