Vector classification of SMD images

The automated visual inspection of surface-mounted devices (SMD) requires the correct classification of an image as either "component present" or "component absent." The inspection system must allow the classification to be fast and reliable while also assuring that the training of the classifier is simple and not time consuming. The traditional sequential approach to classifying images, while simple to implement, presents some disadvantages, including an increase in false alarm (type I) errors and the need for a time-consuming training phase. The method presented in this paper seeks to reduce classification errors by using a vector approach for the inspection of components. An experiment using real SMD inspection data is conducted to validate the performance of the proposed vector classifier. The results of the experiment support the hypothesis that the vector-based inspection approach renders fewer errors than the sequential inspection approach.