• Àüü
  • ÀüÀÚ/Àü±â
  • Åë½Å
  • ÄÄÇ»ÅÍ
´Ý±â

»çÀÌÆ®¸Ê

Loading..

Please wait....

±¹³» ÇÐȸÁö

Ȩ Ȩ > ¿¬±¸¹®Çå > ±¹³» ÇÐȸÁö > µ¥ÀÌÅͺ£À̽º ¿¬±¸È¸Áö(SIGDB)

µ¥ÀÌÅͺ£À̽º ¿¬±¸È¸Áö(SIGDB)

Current Result Document : 11 / 11

ÇѱÛÁ¦¸ñ(Korean Title) ½ÃÁ¤°Å¸® ÆǺ°À» À§ÇÑ CCTV À̹ÌÁöÀÇ Ä­Åõ ¾î ÆÐÅÏ¿¡ ±â¹ÝÇÑ °èÃþÀû Ŭ·¯½ºÅ͸µ ¸ðµ¨
¿µ¹®Á¦¸ñ(English Title) A Hierarchical Clustering Model Based on Contour pattern from CCTV image for Estimating Visibility Distance
ÀúÀÚ(Author) ¾Ë¾Æ½º¸À ¿Ã¶ó   ¹«ÇÔ¸¶µå ŸÀÌ¾ßºê ¾Ë¸®   È«ºÀÈñ   Asmat Ullah   Muhammad Tayyab Ali   Bonghee Hong  
¿ø¹®¼ö·Ïó(Citation) VOL 36 NO. 03 PP. 0137 ~ 0152 (2020. 12)
Çѱ۳»¿ë
(Korean Abstract)
Àú½ÃÁ¤ »óÅ´ ¾È°³°¡ £¾îÁú¼ö·Ï Àΰ£ÀÇ °¡½Ã°Å¸®°¡ ³·¾ÆÁö´Â ±â»óÇö»óÀÌ´Ù. µµ·Î»óÀÇ ±³Åë ¾ÈÀüÀ» À§ÇØ ¾È°³ ¹ß»ý ½Ã ¿îÀü ÀÚ¿¡°Ô °¡½Ã¼º ¼öÁØÀ» º¸¿©ÁÖ´Â ½ÃÁ¤°Å¸®¸¦ ÃßÁ¤ÇÏ´Â ÇнÀ ¸ðµ¨ÀÌ ÇÊ¿äÇÏ´Ù. ÀÌ ³í¹®Àº ½Ç½Ã°£À¸·Î ¿îÀüÀÚ¿¡°Ô ½ÃÁ¤°Å¸®¸¦ º¸¿©ÁÙ ¼ö Àִ Ŭ·¯½ºÅ͸µ ÇнÀ ¸ðµ¨À» Á¦½ÃÇÑ´Ù. µµ·Îº¯ÀÇ CCTV À̹ÌÁö·Î ºÎÅÍ ÃßÃâµÈ µ¥ÀÌÅÍ ÆÐÅÏÀ» Ŭ·¯½ºÅÍ·Î ºÐ·ùÇÏ¿© ½ÃÁ¤°Å¸® ¸¦ ÆǺ°Çϴ Ŭ·¯½ºÅ͸µ ÇнÀ ¸ðµ¨À» Á¦¾ÈÇÑ´Ù. ¸ÕÀú ÆÐÅÏ ÃßÃâÀ» À§ÇØ CCTV ¿µ»óÀ¸·Î ºÎÅÍ °ü½É ¿µ¿ª(ROI)ÀÌ ¼±ÅÃµÇ°í °¢ ROI ÀÇ µ¥ÀÌÅÍ ÆÐÅÏÀÌ ÃßÃâµÈ´Ù. °¢ ROI¿¡¼­ ÃßÃâµÈ µ¥ÀÌÅÍ ÆÐÅÏÀº ¼± ¶Ç´Â »ç¹°ÀÇ À±°û¼± °¹¼ö·Î Ç¥ÇöµÈ´Ù. ÃßÃâµÈ µ¥ÀÌÅÍ ÆÐÅÏ¿¡ ´ëÇÑ °èÃþÀû Ŭ·¯½ºÅÍÀ» ±¸¼ºÇÏ¿© ¼­·Î ´Ù¸¥ ½ÃÁ¤°Å¸® ¼öÁØÀÌ ÆǺ°µÈ´Ù. Ŭ·¯½ºÅÍÀÇ À¯»çµµ ÃøÁ¤À» À§ÇØ °Å¸® À¯»çµµ ÃøÁ¤°ú µ¿Àû ½Ã°£ ¿Ö°î(DTW)À» »ç¿ëÇÏ¿© °èÃþÀû Ŭ·¯½ºÅ͸µÀ» ±¸ÇöÇÏ¿´´Ù. ±×¸®°í ¹ÖÄÚÇÁ½ºÅ°´Â ºñ±³ ¸ñÀûÀ» À§ÇÑ º¸Á¶ ¼ö´ÜÀ¸·Î »ç¿ëÇÏ¿´´Ù. ÀÌ ³í¹®ÀÇ Á¦¾È ¹æ¹ýÀº ƯÁ¤ ÀÓ°è°ª¿¡ µ¶¸³ÀûÀÌ¸ç ´Ù¾çÇÑ ±â»ó Á¶°Ç¿¡¼­ Àß ¼öÇàµÇ´Â »ó´ëÀû ½ÃÁ¤°Å¸® ÃßÁ¤ ¹æ¹ýÀÌ´Ù. ±âÁ¸ µö·¯´× Á¢±Ù ¹æ¹ý°ú ºñ±³ ½ÇÇèÀ» ¼öÇàÇÏ¿´°í ½ÃÁ¤°Å¸® ÆǺ° Á¤È®µµ´Â ¾à 90%ÀÌ´Ù.
¿µ¹®³»¿ë
(English Abstract)
Low visibility condition is a meteorological phenomenon in which the human visibility distance decreases as the fog level increases. To make sure the safety of road traffic, a learning model of estimating visibility distance should be developed, which shows the actual visibility level to the vehicle drivers during foggy weather. In this paper, we have proposed a clustering learning model that provides the visibility level to the drivers in real-time. We use the real CCTV camera images on the roadside and extract a data pattern from it, and then we classify that pattern into different visibility levels by using hierarchical clustering. First, for pattern extraction, we select the optimal regions of interest (ROI) in a CCTV image and then extract features from each ROI. The features we extract are the "number of lines" and the "number of contours" detected in each ROI. Finally, we use hierarchical clustering to classify that data-pattern into different visibility levels. We have implemented hierarchical clustering using two similarity distance metrics, dynamic time warping (DTW) as our main similarity distance metric, and Minkowski is supplementary for comparison purposes. In this paper, we provide a relative visibility estimation method, which is independent of using threshold and performs well in diverse conditions. We have compared our approach to a deep learning method, and it has an overall accuracy of around 90%.
Å°¿öµå(Keyword) µ¥ÀÌÅÍ ºÐ¼®   µ¥ÀÌÅÍ ÆÐÅÏ   °èÃþÀû Ŭ·¯½ºÅ͸µ   Ŭ·¯½ºÅÍ   µ¥ÀÌÅÍ ºÐ·ù   data analysis   data pattern   hierarchical clustering   cluster   data classification  
ÆÄÀÏ÷ºÎ PDF ´Ù¿î·Îµå