Detecting Star Cracks in Topography Images of Specular Back Surfaces of Structured Wafers

In this work, we demonstrate our approach of automatically detecting star cracks in topography images of structured wafers. In these images, we can identify three-dimensional structures typically originating from replicas of chip structures or from grinding artifacts such as grinding grooves or comets. Yet, also defects such as star cracks have a three-dimensional structure and are therefore visible in those images too. To detect the regions containing star cracks, we develop a pattern recognition algorithm consisting of two cascaded classifiers. With this algorithm, we reach a recall rate of 96.2% and a precision rate of 99.3% of the class star crack, respectively.