Extraction and Detection of Singular Manufacturing Features using KAZE Descriptors

Manufacturing features are geometric volumes whose manufacturability is dependent on the type of machining operation, material composition, required tolerances, and available tooling. The automatic recognition of manufacturing features on parts is challenging as many manufacturing features exhibit high degrees of similarity and interact in complex ways that are difficult for traditional methods to identify. As a result, recognition is generally performed manually by human planners who are prone to error. This work uses KAZE descriptors to create a computer vision algorithm capable of automatically extracting and detecting manufacturing features on digital CAD models using a reference image library. Results indicate that the proposed computer vision algorithm is able to recognize manufacturing features in a way that is accurate, precise, and reliable. This is demonstrated through the detection of a singular manufacturing feature located on a part. Once identified, manufacturing features can be used to simplify downstream planning by mapping part characteristics to machining capabilities.