Saliency Map Based Defect Alarm System for FPC Roll-to-Roll Process

Defect rate in flexible printed circuit(FPC) industry is not so good. This paper propose a system and an algorithm to monitor appearance defects in FPC roll-to-roll process. This algorithm is based on selective attention mechanism computed by saliency map. Edge density is introduced as feature to compute the saliency map. Object’s distribution map is defined to help locate all the known objects in FPC. According to the particularity of roll-to-roll process, lateral difference operator is presented to highlight the defects. This operator’s algorithm involved the edge density maps and object’s distribution maps. Experimental data confirm the validity of the proposed system and algorithm.