STATISTICAL LEARNING-BASED MODE SELECTION FOR MULTI-MODE INSPECTION

摘要

Methods and systems for selecting mode(s) for inspection of specimens are provided. One method includes statistically predicting if data points in a set correspond to defects or nuisances on a specimen. The data points include attribute(s) determined for discrete locations on the specimen from output generated by two or more modes of an inspection system. Events have been detected at the discrete locations with at least one of the modes. The method also includes determining a quantitative measure for each of two or more different combinations of the modes thereby determining different quantitative measures. The quantitative measure for each of the different combinations is responsive to how well one of the combinations detects the defects and minimizes detection of the nuisances. The method further includes selecting one or more of the modes for inspection of specimens of the same type as the specimen based on the determined quantitative measures.