Reflection-area-based feature descriptor for solder joint inspection

This paper presents a new set of features for the machine inspection of solder joints. The present approach is very attractive for industrial applications mainly due to its simplicity in illumination and classification method, efficiency in computation time and robustness to changes in the illumination condition and the pad size. It is remarkable that the present approach requires only one layer of tiered light, and its classification method makes use of only two features so that insufficient, acceptable and excessive solders can be divided into three classes by two straight lines on the two-dimensional feature plane. And it has been verified by experiments that it can acutely discriminate solder joints of different classes. The simplicity, reliability and computational efficiency of the present approach are all owing to the careful selection of a set of features which are derived step by step from well-established physical laws, though only in qualitative sense. Hence this paper demonstrates also the advantage of physics-based pattern classification approaches.