ÀϽà : 2022.05.05~2022.05.06Àå¼Ò : Á¦ÁÖICC
ISSN 2950-9467
ÁÖ°ü±â°ü : Çѱ¹ ºòµ¥ÀÌÅͼºñ½ºÇÐȸ
ÃÖ±Ù ´Ù¾çÇÑ Á¦Á¶¾÷ ºÐ¾ß¿¡¼ µðÁöÅÐÀüȯÀÌ ÁøÇàµÇ°í ÀÖÀ¸¸ç, ÄÜÅÃÆ®Á¦Á¶ ±â¾÷¿¡¼µµ ºòµ¥ÀÌÅÍ ¹× ÀΰøÁö´É ±â
¼úÀÇ µµÀÔÀÌ ÁøÇàµÇ°í ÀÖ´Ù. ÀΰøÁö´É ±â¼úÀ» ÀÌ¿ëÇÑ Ä®¶ó ÄÜÅÃÆ®·»Áî Á¦Á¶ °øÁ¤¿¡¼ ¹ß»ýÇÏ´Â ºÒ·® À¯¹«ÀÇ Á¤È®ÇÑ
ŽÁö ±â¼úÀÌ ºÎ°¢µÇ°í ÀÖ´Ù. º» ³í¹®¿¡¼´Â ÄÜÅÃÆ®·»ÁîÀÇ ºÒ·®À» ŽÁöÇϱâ À§ÇØ ÇʼöÀûÀ¸·Î ¿ä±¸µÇ´Â Ä®¶óÀμâ
ÆÐÅÏÀ» È¿À²ÀûÀ¸·Î ã´Â 2Áß ±¸Á¶ÀÇ CNN ±â¹ýÀ» Á¦½ÃÇÑ´Ù. Á¦½ÃÇÑ ±â¹ýÀ» Àû¿ëÇϱâ À§ÇØ 5¸¸ °³ÀÇ ¿µ»óµ¥ÀÌÅ͸¦
»ç¿ëÇÏ¿© ÀμâÆÐÅÏÀÇ ¿µ¿ªÀ» °ËÃâÇϱâ À§ÇÑ ÇнÀ¸ðµ¨À» ¸¸µé¾ú°í, ÀÌ ÇнÀ¸ðµ¨À» ÀÌ¿ëÇÏ¿© 1,000°³ ¿µ»ó¿¡ ´ëÇÑ
½ÃÇè °á°ú, 0.05 MAE(Mean Absolute Error)¸¦ ¾ò¾ú´Ù. Abstract
Recently, digital transformation is in progress in various manufacturing fields, and the introduction of big data
and artificial intelligence technologies is also in progress in contact manufacturing companies. Accurate detection
of defects occurring in the color contact lens manufacturing process using artificial intelligence technology is
emerging.In this paper, we present a CNN method with a double structure that efficiently finds the color printing
pattern, which is essential to detect contact lens defects.In order to apply the proposed technique, a learning model
was created to detect the print pattern area using 50,000 image data, and the test result for 1,000 images, 0.05
MAE(Mean Absolute Error) was obtained using this learning model.