3D vision based quality inspection with computational intelligence

Purpose Product quality inspection is of importance in manufacturing industries to ensure that low quality or unqualified products are not delivered to the consumer. Human inspection has many limitations such as low accuracy or speed due to factors such as tiredness and boredom. Traditional 2D vision inspection also has limitations of product shape complexity or flexibility. Thus, automated 3D vision inspection is anticipated to meet the requirements of higher applicability. This paper seeks to address these issues. Design/methodology/approach In many product quality inspection problems, geometrical parameters of the industrial parts are commonly used as the basis of quality inspection. Machine vision is widely applied to acquire such kind of parameters. Comparing to traditional 2D vision, 3D vision can acquire 3D coordinates of the object directly, so that the inspection can be accomplished which is difficult to do with 2D vision. As an active vision technique, structure light system (SLS) is applied to acquire the 3D coordinate information of inspected object in this paper. On the basis of point cloud and regression analysis, features relative to quality are defined and extracted as the attributes for the product classification. Three data mining techniques are applied to accomplish the classification in this paper, which include decision trees, artificial neural networks and support vector machine. Findings A new intelligent automated 3D vision quality inspection for assembly lines has been developed, which comprises structure light system (SLS) and data mining approaches such as decision tree, artificial neutral networks and support vector machine. Originality/value The combination of structure light system (SLS) and data mining approaches makes the automated quality inspection available. The proposed system is easy to be implemented and flexible for different types of products.