| 摘要 |
<p>Intended is to detect the defect of an MEMS device highly accurately and efficiently without needing any absolute model image and to classify the kind of the defect precisely. A defect detecting device (100) images a protein chip (35) formed in each die (30) of a wafer (1), at a low magnification for each first division region (71) having the die divided into a plurality, and stores the protein chip as an inspection target image together with an ID for identifying each first division region (71). The device calculates an average brightness value of each pixel of each inspection target image having the corresponding ID, and creates a model image for each first division region (71). The device extracts the difference between the model image and each inspection target image, as a difference image, and then performs a Blob extraction on each difference image to extract the Blob of a predetermined area or more, thereby to decide the presence or absence of a defect. In case the defect is present, moreover, the device images each second division region (72) divided from the first division region (71), at a high magnification, and creates a model image again to extract the Blob thereby to classify the kind of the defect on the basis of the featuring point of the defect.</p> |
