Study on Feature Point Matching for Diverse Workpieces Based on an Improved Speeded Up Robust Features(SURF) Algorithm

Abstract

Owing to variations in the size, shape, moving speed, and direction of different workpieces, robotic arm vision systems exhibit insufficient capability in identifying feature points across diverse work-pieces, resulting in limited flexibility of visual recognition. To address this issue, this study focuses on feature matching methods for multi-workpiece identification, introducing a feature point matching algorithm into the visual module. Specifically, A multi-workpiece recognition feature matching method based on the improved Speeded Up Robust Features (SURF) algorithm is proposed. The method incorporates the extraction of salient regions to reduce the number of mismatched points and adopts the principal component analysis (PCA) dimensionality reduction principle to reduce the dimension of feature point description vectors, thereby shortening the matching time. Furthermore, comparative experiments were conducted against the traditional SURF algorithm. Experimental results demonstrate that, compared to the traditional algorithm, the improved SURF algorithm enhances recognition accuracy by over 95% and improves matching speed by nearly 50%, contributing to the further optimization of the SURF algorithm’s matching performance. Workpiece recognition experiments were conducted using a KUKA robotic arm for validation, which also confirms that the proposed improved SURF algorithm can meet the requirements of workpiece recognition in practical working environments.