A multi-workpieces recognition algorithm based on shape-SVM learning model

Abstract

In order to achieve the goal of the robot which has capability of learning and grasping the predetermined workpieces actively on the assembly line, a multi-workpieces recognition algorithm based on SSLM (shape-SVM learning model) is proposed. In contrast to traditional feature model which requires great effort to establish model library for the specific workpiece, SSLM is much easier to train and learn, even when it is applied to different object across complex environment in our experiments, the excellent performance can be achieved by almost same settings of SSLM. To make the training recognition algorithm free from the influence of workpieces dimension and rotation, SVH (Shape vector histogram) is created to express and wrap the contour features, and then SVM is adopted to complete the training and prediction of workpiece type in this paper. More than 2000 workpieces are identified in term of proposed algorithm which has an accuracy rate of 98%.