Keyframe Extraction and Process Recognition Method for Assembly Operation Based on Density Clustering

Abstract

A keyframe extraction and process recognition method for assembly operation is proposed based on density clustering to solve the problems of data redundancy and difficulty in obtaining valid data frames from the process of continuous assembly operation. A standard operation gesture set including dynamic and static action was constructed through assembly operation decomposing. Finger feature variables and gesture comprehensive feature quantized function were defined according to finger joint structure. Based on searching for local extreme points of the function, density clustering method was used to extract the keyframes of the assembly operation sequence to eliminate the redundant data. Finally, the support vector machine algorithm model and the Levinstein distance were determined to complete the keyframe recognition and assembly operation matching. A case study showed that the method presented can effectively discretize the assembly operation sequence, remove approximately 84% of redundant data frames, and achieve a comprehensive recognition rate of 92%.