Selection of task specific indicator functions in a multiscale measurement system

Modern inspection tasks often require high resolution and fast measurement times in combination with rather large object fields. Multiscale measurement systems offer a solution to balance these criteria. A crucial task is to find the indicators that give a hint to unresolved defects on the specimen in the actual sensor scale. Finer scaled sensors are necessary to further classify these indicators. This contribution focuses on the automatic selection of relevant indicator detection functions specific to a given task. At each scale, sensor data is simulated using different possible defects models. This simulated sensor data is processed using multiple indicator functions (with automatic parameter optimization) to check if a particular indicator function is suitable for the given task. Algorithm evaluation functions are used to verify the reliability of the indicator function. We present results concerning the optimization of indicators for the detection of shape errors, pointlike errors and scratches on microlens arrays in a multiscale measurement system using video microscopy and confocal microscopy.