Abstract
4Â÷ »ê¾÷Çõ¸í°ú ÇÔ²² µðÁöÅÐ Àüȯ(Digital Transformation, DX) ±â¼úÀÌ Áß¿äÇØÁö°í ÀÖ´Ù. ÀÌ¿Í ÇÔ²² Àΰø
Áö´ÉÀ» ÅëÇÑ »ý»ê°øÁ¤¿¡¼ÀÇ ºÒ·® °ËÃâ ¹× ºÐ·ù¿¡ ´ëÇÑ ¿¬±¸°¡ È°¹ßÈ÷ ÀÌ·ç¾îÁö°í ÀÖ´Ù. º» ³í¹®¿¡¼´Â ´Ù¾çÇÑ
CNN ¸ðµ¨À» »ç¿ëÇÏ¿© Ä÷¯ ÄÜÅÃÆ®·»Áî »ý»ê°øÁ¤¿¡¼ ¹ß»ýÇÏ´Â ºÒ·® °ËÃâÀ» È¿°úÀûÀ¸·Î ¼öÇàÇÏ´Â ¸ðµ¨À» ¼±
Á¤ÇÏ°íÀÚ Çϸç, À̸¦ ÅëÇØ »ý»ê ¹× Ç°ÁúÀÇ Çâ»óÀ» ÀÌ·ç¾î, ÀÚ¿øÀÇ ³¶ºñ¿Í ¼ÒºñÀÚÀÇ ¾ÈÀüÀ» È®º¸ÇÏ°íÀÚ ÇÑ´Ù.
À̸¦ À§ÇØ Ä÷¯ ÄÜÅÃÆ®·»Áî ¿µ»ó¿¡ ´ëÇÑ Àüó¸®¿Í Áõ°À» ÅëÇØ ÇнÀ ¹× °ËÁõ µ¥ÀÌÅ͸¦ »ý¼ºÇÏ¿´À¸¸ç, RGB
¹× HSV ä³Î ¿µ»ó¿¡ ´ëÇØ ResNet101, GoogLeNet V2, GoogLeNet V4, DenseNet121, MobileNetÀÇ
CNN ±â¹ýÀ» È°¿ëÇÏ¿© RGB¿Í HSV ä³Îº°·Î ºÒ·® ŽÁöÀ² ºñ±³ ºÐ¼®ÇÏ¿´´Ù. À§ ¸ðµ¨ÀÇ Á¤È®µµ´Â ¼ø¼´ë·Î
°¢°¢ 89.74%, 84.46%, 95.43%, 82.80%, 89.74%·Î, RGB ä³ÎÀÇ GoogLeNet V4°¡ °¡Àå ³ôÀº ºÒ·® °ËÃâ
Á¤È®µµ¸¦ ¾ò¾úÀ¸¸ç, ´ëºÎºÐÀÇ ¸ðµ¨¿¡¼ RGB ä³ÎÀÌ HSV ä³Îº¸´Ù ´õ ÁÁÀº °á°ú¸¦ ¾ò¾î³¿À» ¾Ë ¼ö ÀÖ¾ú´Ù.
The importance of Digital Transformation (DX) technology has increased with the Fourth Industrial
Revolution. At the same time, research on defect detection and classification in the production process
through artificial intelligence has been actively applied. In this paper, we select a model that effectively
detects defects that occur in the production process of color contact lenses using various models,
secure reducing resource waste and consumer safety by improving production and quality. For this
purpose, data for training and validation were generated through preprocessing and augmentation of
color contact lens images, using CNN technologies such as ResNet101, GoogLeNet V2, GoogLeNet V4,
DenseNet121, MobileNet compared and analyzed the defect detection rate for each RGB channel and
HSV channel. The accuracies of the above models are 89.74%, 84.46%, 95.43%, 82.80%, and 89.74%
respectively, with GoogLeNet V4 on the RGB channel having the highest defect detection accuracy, and
in most models, the RGB channel is higher than the HSV channel.