Computer aided design (CAD) model search and retrieval using frequency domain file conversion

Abstract

The computer aided design (CAD) files are often tagged manually or searched and retrieved from a database using machine learning algorithms. There has been significant interest in developing capabilities of searching CAD files based on shape of the object. The present work demonstrates a method that transforms the CAD files to frequency domain representations and generates a set of fingerprints from the spectrogram to enable a search function. The fingerprints can be adjusted to allow finding either an exact match or a broader range of approximate/partial matches with similar features. A matching index is developed based on the similarity between the designs. The search does not require manual tagging of the files and is purely based on the geometry of the object. The success in comparing two files based on the object geometry is an important capability that can allow authentication of files in addition to conducting the search function.