High-throughput shape classification using support vector machine

Identification of defective objects is a crucial problem in assembly line industries. Several machine learning techniques along with image processing have been proposed to solve this problem. However such techniques are limited to low flow rate lines and hence constrain the productivity of the line. In this paper, a method is proposed to identify small defective objects in a high-throughput line. A machine learning technique called Support Vector Machines (SVM) along with a simple set of interrupt sensors (such as infra-red optical sensors) is used to identify such objects. The method demonstrates nearly the same accuracy as that achieved using conventional image processing techniques.