| 摘要 |
PURPOSE:To obtain a high recognition rate by allowing a character recognizing part to execute character recognition based upon a neural network. CONSTITUTION:An image input part 1 enters a wafer identification(ID) character by a camera 11 as input data 21 and a binarizing part 12 binarizes the data 21 received from the input part 1. Binarized data are received by a scribe line/ bonding pad removing part 13 to remove a bonding pad and a scribe line when a noise included in an image is generated. The image is thickened until the dots of the character are superposed, the image is thinned until only the bonding pad and the scribe line are erased and then the image is thickened again. The processed image is transferred to a segmenting part 14, which segments corrected data 23 in each character. A character recognizing part 3 receives the segmented character data 24, recognizes the data 24 by a neural network part 15 and returns the recognized result 25. |
