³í¹®¸í |
REBA Æò°¡ µµ±¸ ±â¹Ý ÀÛ¾÷ÀÚÀÇ ÀÚ¼¼ ºÐ¼® ÀÚµ¿È Á¦¾È / Automated Posture Analysis Proposal Based on REBA Evaluation Tool |
¼ö·Ï»çÇ× |
´ëÇÑ°ÇÃàÇÐȸ³í¹®Áý, Vol.40 No.3 (2024-03) |
ÆäÀÌÁö |
½ÃÀÛÆäÀÌÁö(265) ÃÑÆäÀÌÁö(8) |
ÁÖÁ¦¾î |
Çൿ ÀνÄ; Á÷¾÷°ü·Ã ±Ù°ñ°Ý°è Áúȯ; ÀΰøÁö´É; Ƽóºí ¸Ó½Å ; Action Recognition; Work-related Musculo-skeletal Disorders; Artificial Intelligence; Teachable Machine |
¿ä¾à1 |
°Ç¼³¾÷°è¿¡¼´Â ±Ù·ÎÀÚÀÇ ¾ÈÀüÀ» À§ÇØ ÄÄÇ»ÅÍ ºñÀüÀ» È°¿ëÇÑ ÇöÀå°ü¸® ¹× ºÐ¼® ¿¬±¸°¡ Á¡Â÷ Áõ°¡ÇÏ°í ÀÖ´Ù. ¾ÈÀü¸ð Âø¿ë, ¾ÈÀü°í¸® ü°á È®ÀÎ, ÁßÀåºñ °Åµ¿ ÀÚµ¿ ÀÎ½Ä µî ÄÄÇ»ÅÍ ºñÀüÀ» È°¿ëÇÏ°í ÀÖ´Ù. ±×·¯³ª ±Ù°ñ°Ý°è ÁúȯÀÌ ¹ß»ýÇÏ´Â ÀÚ¼¼¿¡ ´ëÇÑ Á÷Á¢ÀûÀÎ ¿¬±¸´Â »ó´ëÀûÀ¸·Î ¹ÌÈíÇÏ´Ù. °Ç¼³ÇöÀåÀº °øÁ¤º°·Î °¢±â ´Ù¸¥ Àü¹®Àη°ú Àåºñ°¡ ÅõÀԵǴ ³ëµ¿Áý¾àÀû »ê¾÷ÀÌ´Ù. °Ç¼³±Ù·ÎÀÚÀÇ ±Ù°ñ°Ý°è Áúȯ ¹ß»ýÀ» ÁÙÀÌ°í ±Ù·ÎÀÚÀÇ ¾ÈÀüÀ» º¸È£Çϱâ À§Çؼ´Â Áö¼ÓÀûÀÎ °ü¸®¿Í ¸ð´ÏÅ͸µÀÌ Áß¿äÇÏ´Ù. ±×·¯³ª °ø»çÇöÀåÀº ±Ô¸ð°¡ Å©°í °øÁ¤¿¡ µû¶ó ¾÷¹«°¡ ´Þ¶ó °ü¸®ÀÚµéÀÌ °ü¸®Çϱ⠾î·Æ´Ù. µû¶ó¼ º» ¿¬±¸¿¡¼´Â REBA¸¦ ±â¹ÝÀ¸·Î À§ÇèÀÚ¼¼¸¦ Á¤ÀÇÇÏ°í PosenetÀ» ÀÌ¿ëÇÏ¿© °üÀýÀÇ ÇÙ½É Æ÷ÀÎÆ®¸¦ ÃßÃâÇÏ¿´´Ù. À§ÀÇ ÀڷḦ ¹ÙÅÁÀ¸·Î Teachable MachineÀ» »ç¿ëÇÏ´Â ÀÛ¾÷ÀÚÀÇ Åµµ¸¦ ºÐ·ùÇÏ´Â ¸ðµ¨À» Á¦¾ÈÇÏ¿´´Ù. Á¦¾ÈµÈ ¸ðµ¨Àº °¢°¢ÀÇ À§Çè ÀÚ¼¼¸¦ ÀνÄÇÏ´Â Á¤È®µµ°¡ ³ôÀº °ÍÀ¸·Î ³ªÅ¸³µ´Ù. |
¿ä¾à2 |
In the construction industry, there is a gradual increase in the application of computer vision for field management and safety analysis of workers. Computer vision is employed for tasks like monitoring the use of safety helmets, verifying the fastening of safety rings, and automatically recognizing the behavior of heavy equipment. However, research specifically addressing the postures leading to musculoskeletal disorders is relatively limited. The construction site, being labor-intensive and involving various professionals and equipment in each process, requires continuous management and monitoring to minimize musculoskeletal diseases among workers and ensure their safety. Managing such a large construction site with diverse tasks for each process poses challenges for effective oversight. In this study, risk postures were defined based on REBA, and key joint points were identified using Posenet. Using this data, a model was developed to classify workers' postures using Teachable Machine, demonstrating high accuracy in recognizing different risk postures |