| ³í¹®¸í |
MEMS ¶óÀÌ´Ù ¼¾¼¸¦ Ȱ¿ëÇÑ ½ÉÃþÇнÀ ±â¹Ý Á¶Àûº®Ã¼ °áÇÔ ÀÎ½Ä ±â¼ú / Deep Learning based Masonry Wall Defect Classification using a MEMS LiDAR Sensor |
| ÀúÀÚ¸í |
Ȳ¿µ¼(Hwang, Yeongseo) ; ¹Ú±ÙÇü(Park, Gunhyung) ; ¾ç°Çõ(Yang, Kanghyeok) |
| ¼ö·Ï»çÇ× |
´ëÇѰÇÃàÇÐȸ³í¹®Áý, Vol.39 No.1 (2023-01) |
| ÆäÀÌÁö |
½ÃÀÛÆäÀÌÁö(313) ÃÑÆäÀÌÁö(8) |
| ÁÖÁ¦¾î |
½ÉÃþÇнÀ; MEMS ¶óÀÌ´Ù; 3D ·¹ÀÌÀú ½ºÄ³³Ê; Á¶Àûº®Ã¼; °áÇÔ ÀÎ½Ä ; Deep Learning; MEMS LiDAR; 3D Laser Scanner; Masonry Wall; Defect Classification |
| ¿ä¾à1 |
°ÇÃ๰ÀÇ À¯Áö°ü¸® ¹× ¾ÈÀüÁ¡°ËÀº ´ëºÎºÐ Á¡°ËÀÚÀÇ À°¾ÈÀ¸·Î ÁøÇàÇÏ¿© ¸¹Àº ½Ã°£°ú ÀηÂÀÌ ¼Ò¸ðµÈ´Ù´Â ¹®Á¦Á¡ÀÌ ÀÖ´Ù. À̸¦ º¸¿ÏÇϱâ À§ÇØ ¿µ»ó󸮱â¼ú ¹× ÀΰøÁö´ÉÀ» Ȱ¿ëÇÑ °áÇÔ ÀÎ½Ä ±â¼ú °³¹ßÀÌ È°¹ßÇÏ°Ô ÁøÇàµÇ°í ÀÖ´Ù. ÇÏÁö¸¸ ±âÁ¸ÀÇ ¿µ»óó¸® ±â¹ýÀº Ä«¸Þ¶ó¸¦ ÅëÇØ ¾òÀº À̹ÌÁö¸¦ ºÐ¼®ÇÏ´Â ¹æ½ÄÀ¸·Î ÁÖº¯ ȯ°æ¿¡ µû¶ó ¼º´ÉÀÌ º¯ÇÏ´Â ÇѰ谡 ÀÖ´Ù. ÃÖ±Ù À̸¦ ÇØ°áÇϱâ À§ÇØ 3D ·¹ÀÌÀú ½ºÄ³´× ¼¾¼¸¦ ÀÌ¿ëÇÑ °áÇÔÀÎ½Ä ¹æ¹ýÀ» °³¹ßÇÏ¿´À¸³ª ÀåÄ¡ÀÇ °¡°ÝÀÌ ºñ½Î Ȱ¹ßÇÑ È°¿ëÀÌ ¾î·Æ´Ù´Â ´ÜÁ¡ÀÌ ÀÖ´Ù. ÀÌ¿¡ º» ¿¬±¸´Â ±âÁ¸ ½ºÄ³´× ÀåÄ¡º¸´Ù °¡°ÝÀÌ Àú·ÅÇÏ°í ½Å·ÚÇÒ¸¸ÇÑ ¼º´ÉÀ» º¸À̰í ÀÖ´Â MEMS ¶óÀÌ´Ù ¼¾¼¸¦ ÀÌ¿ëÇØ Á¶Àûº®Ã¼ÀÇ °áÇÔÀ» ÀνÄÇÒ ¼ö ÀÖ´Â ±â¼úÀ» °³¹ßÇÏ¿´´Ù. ÇØ´ç ¿¬±¸´Â Á¶Àûº®Ã¼¸¦ ´ë»óÀ¸·Î ÇÏ¿´À¸¸ç, ½ÇÇè½Ç ȯ°æ¿¡¼ ¿©·¯ Á¾·ùÀÇ °áÇÔÀ» °¡Áø ½ÃÇèü¸¦ Á¦ÀÛÇÏ¿© µ¥ÀÌÅ͸¦ ȹµæÇÏ¿´´Ù. Á¶Àûº®Ã¼ °áÇÑ ÀÎ½Ä ¹æ¹ýÀ¸·Î ÀΰøÁö´ÉÀ» Ȱ¿ëÇÑ ¿¬±¸¿¡¼ ¸¹ÀÌ »ç¿ëÇϰí ÀÖ´Â ResNet-50°ú VGG16 ¸ðµ¨À» »ç¿ëÇÏ¿© °áÇÔÀ» ÀνÄÇÏ¿´À¸¸ç, ¼º´ÉÆò°¡ °á°ú ResNet-50Àº 98.75%, VGG16Àº 96.88%ÀÇ Á¤È®µµ¸¦ º¸¿©ÁÖ¾ú´Ù. ÇØ´ç ¿¬±¸ °á°ú´Â ¸ð¹ÙÀÏ 3D ·¹ÀÌÀú ½ºÄ³´× ÀåÄ¡¿Í °áÇÕÇÏ¿© Á¶Àûº®Ã¼ÀÇ ½Ç½Ã°£ °áÇÔ ÀÎ½Ä ±â¼ú °³¹ß¿¡ Ȱ¿ëµÉ ¼ö ÀÖÀ» °ÍÀ¸·Î ÆÇ´ÜµÈ´Ù. |
| ¿ä¾à2 |
Most of the maintenance and safety inspections of buildings are performed with visual assessment of the inspector, which consumes a lot of time and cost. With the development of computer vision and digital technologies such as 3D Laser scanners, automatic defect recognition using image processing and artificial intelligence has been widely studied. Current approach is largely relying on the image obtained from the camera and the recognition performance could be varied depending on the surrounding environment. Recently, studies using 3D Laser scanner are being conducted to solve these problems. However, terrestrial laser scanners are expensive, so it is difficult to apply at the construction site. Therefore, this study proposed a method that can recognize masonry wall defects using a Microelectromechanical systems based Light Detection and Ranging sensor that having much lower price and reliable performance. This study was performed using masonry wall structures and data were collected from samples having various types of defects in a laboratory environment. Masonry wall defects were recognized using ResNet-50 and VGG16 models, which are widely used in previous studies. As a result of the classification, ResNet-50 and VGG16 achieved 98.75% and 96.88% accuracy, respectively. The results of this study can be utilized in the development of real-time defect recognition method for a masonry wall at construction sites. |