| ¼ö·Ï»çÇ× |
Çѱ¹±¸Á¶¹°Áø´ÜÀ¯Áö°ü¸®°øÇÐȸ ³í¹®Áý , Vol.29 No.5(2025-10) |
| ÁÖÁ¦¾î |
¹«Àαâ; µå·Ð; ÀÔ¸é Á¤»ç¿µ»ó; ±Õ¿; ±³·®; ¿Ü°üÁ¶»ç¸Áµµ ; UAV; Drone; Facade orthomosaic; Crack; Bridge; Exterior damage map |
| ¿ä¾à1 |
µå·ÐÀ» Ȱ¿ëÇÑ ±³·® Á¡°ËÀº µ¥ÀÌÅÍ ÃëµæÀÇ È¿À²¼ºÀ» ³ô¿´À¸³ª, ¿µ»ó ºÐ¼®Àº ÁÖ·Î ±Õ¿ÀÇ À§Ä¡¸¦ ÆÄ¾ÇÇÏ´Â ´Ü¼ø ŽÁö¿¡ ¸Ó¹«¸£´Â °æ¿ì°¡ ¸¹¾Ò´Ù. ƯÈ÷, ŽÁöµÈ ±Õ¿ÀÇ Æø°ú ±æÀ̸¦ Á¤·®ÀûÀ¸·Î ºÐ¼®ÇÏ°í ±× ½Å·Úµµ¸¦ ÇöÀå ½ÇÃø µ¥ÀÌÅÍ¿Í ºñ±³ °ËÁõÇÏ´Â ¿¬±¸´Â ºÎÁ·ÇÏ´Ù. º» ¿¬±¸¿¡¼´Â ÀÌ·¯ÇÑ ÇѰ踦 ±Øº¹ÇϰíÀÚ, µå·Ð ÀÔ¸é Á¤»ç¿µ»óÀ» ±â¹ÝÀ¸·Î 0.1 mm ¼öÁØÀÇ ¹Ì¼¼ ±Õ¿À» ÀÚµ¿À¸·Î ŽÁö ¹× Á¤·®ÈÇϰí, ±× °á°ú¸¦ ¹ÙÅÁÀ¸·Î ¿Ü°üÁ¶»ç¸Áµµ¸¦ ÀÛ¼ºÇÏ´Â ÅëÇÕ ¿öÅ©Ç÷ο츦 Á¦¾ÈÇÏ°í °ËÁõÇÏ¿´´Ù. À̸¦ À§ÇØ ÃÊÇØ»óÈ ¸ðµ¨·Î ¿µ»ó ǰÁúÀ» Çâ»ó½ÃŲ ÈÄ, Àǹ̷ÐÀû ºÐÇÒ¸ðµ¨À» Àû¿ëÇÏ¿© ±Õ¿À» ŽÁöÇßÀ¸¸ç, ŽÁöµÈ ±Õ¿Àº º¤ÅÍÈ ÈÄó¸® °úÁ¤À» ÅëÇØ ±æÀÌ¿Í Æø Á¤º¸¸¦ »êÃâÇÏ¿´´Ù. »êÃâµÈ °á°úÀÇ ½Å·Ú¼ºÀº ÇöÀåÀü¹®°¡°¡ Á÷Á¢ ÃøÁ¤ÇÑ ½ÇÃø°ª°ú ºñ±³ÇÏ¿© °ËÁõµÇ¾ú´Ù. ±× °á°ú, ±Õ¿ ŽÁö Á¤¹Ðµµ´Â 90%¿¡ ´ÞÇßÀ¸¸ç, AI°¡ ÃøÁ¤ÇÑ ±Õ¿ ±æÀÌ´Â Àü¹®°¡ °èÃø°ª°ú Æò±Õ 91%ÀÇ ³ôÀº À¯»çµµ¸¦ º¸¿´´Ù. ±Õ¿ Æø ÃøÁ¤ ¿ÀÂ÷´Â Æò±Õ 0.0175 mm·Î ¸Å¿ì ³·°Ô ºÐ¼®µÇ¾ú´Ù. º» ¿¬±¸´Â µå·Ð À̹ÌÁö·ÎºÎÅÍ ½Å·Úµµ ³ôÀº Á¤·® µ¥ÀÌÅ͸¦ ÀÚµ¿À¸·Î ÃßÃâÇÏ¿© ÃÖÁ¾ ¿Ü°üÁ¶»ç¸Áµµ ÀÛ¼º¿¡ Ȱ¿ëÇÏ´Â ±â¼úÀÇ ½Ç¿ë¼ºÀ» ÀÔÁõÇÔÀ¸·Î½á, ±³·® ¾ÈÀü Á¡°ËÀÇ È¿À²¼º°ú °´°ü¼ºÀ» Çâ»ó½ÃŰ´Â µ¥ ±â¿©ÇÒ ¼ö ÀÖÀ½À» º¸¿´´Ù. |
| ¿ä¾à2 |
While Uncrewed Aerial Vehicle inspections enhance data acquisition, subsequent analysis is often limited to simple crack detection, lacking quantitative analysis and validation against on-site measurements. To address this gap, this study proposes and validates an integrated workflow that automatically detects and quantifies micro-cracks as fine as 0.1 mm from facade orthomosaics to create an exterior damage map. The methodology enhances image quality via a super-resolution model, detects cracks using semantic segmentation, and calculates their length and width through vectorization post-processing. The reliability of these quantitative results was validated against on-site measurements performed by an expert. The results demonstrated a high crack detection precision of 90%. The AI-estimated crack lengths showed an average similarity of 91% to expert measurements, and the crack width estimation yielded a low mean error of 0.0175 mm. This study proves the practicality of using automated quantitative analysis from drone imagery to create reliable exterior damage maps, thereby enhancing the efficiency and objectivity of bridge safety inspections. |