A low-cost imaging system capable of recognizing and counting semiconductor good dice is designedand is proven experimentally to be effective. The system utilizes a commercial scanner with the leastmodification of its inner optics by a beamsplitter-insertion method. Patterns on specular die surfacestherefore can be grabbed clearly, which results in the marks on defective dice being distinguishable.Also, a mask algorithm is developed to count the good dice on an adhesive tape, according to thegray-level histogram and the converted binary picture of the scanned image. The built-in softwareassociated with the scanner, which is reliable and user-friendly, can also be incorporated into thedesignated algorithm without paying an additional price for system development.
展开▼