A New Inspection Method of Chip Component Type Based on Moment Features

Abstract

In order to automatically inspect chip components with the placement machine vision system, a new method is introduced to inspect the type of a chip component. This work studies the moment features selection for chip components classification. First, a new signal to noise ratio formula is proposed to exclude the unrelated moment features with classification, and then a support vector machine is applied to validate the classification performance of the remaining moment features. Then the correlation analysis and the sensitivity analysis are implemented to filter out the redundant moment features. Finally, a voting strategy of one to one is used to promote the support vector machine to obtain a multiclass classifier. Six types of components are inspected for the experiments, and the results show the effectiveness of the type inspection method.