Á¤º¸°úÇÐȸ ³í¹®Áö B : ¼ÒÇÁÆ®¿þ¾î ¹× ÀÀ¿ë
Current Result Document :
ÇѱÛÁ¦¸ñ(Korean Title) |
¹°Ã¼ÀνÄÀ» À§ÇÑ °èÃþÀû ±¸Á¶ ÆÐÅÏ ±â¹ÝÀÇ ÀÌÁø ±â¼úÀÚ |
¿µ¹®Á¦¸ñ(English Title) |
Hierarchical Structure Pattern based Binary Descriptor for Object Recognition |
ÀúÀÚ(Author) |
±èÀμö
¼º¸íö
±è´ëÁø
Insu Kim
Myungchul Sung
Daijin Kim
|
¿ø¹®¼ö·Ïó(Citation) |
VOL 41 NO. 06 PP. 0440 ~ 0447 (2014. 06) |
Çѱ۳»¿ë (Korean Abstract) |
ÃÖ±Ù ¸ð¹ÙÀÏ ½ÃÀåÀÌ Áõ´ëÇÏ¸é¼ È¸Àü, Å©±â, ¾îÆÄÀÎ º¯È¯¿¡ ´ëÇØ °ÀÎÇÏ¸é¼ µ¿½Ã¿¡ °í¼Ó 󸮰¡ °¡´ÉÇÑ ±â¼úÀÚ¿¡ ´ëÇÑ ¿ä±¸°¡ Áõ°¡ÇÏ°í ÀÖ´Ù. º» ³í¹®¿¡¼´Â 2Â÷¿ø ¿µ»ó¿¡¼ ¹°Ã¼ ÀÎ½Ä ¹× Æ÷Áî ÃßÁ¤À» À§ÇÑ ¿µ»ó Ç¥Çö ±â¼úÀÎ °èÃþÀû ±¸Á¶ ÆÐÅÏ¿¡ ±â¹ÝÇÑ »õ·Î¿î ÀÌÁø ±â¼úÀÚ(Binary descriptor)¸¦ Á¦¾ÈÇÑ´Ù. ÃÖ±Ù Á¦¾ÈµÈ ÀÌÁø ±â¼úÀÚ FREAK, BRISK µîÀÇ ¿¬±¸´Â ±âÁ¸ÀÇ SIFT-likeÇÑ ±â¼úÀÚÀÇ ÀÎ½Ä ¼º´ÉÀ» À¯ÁöÇÏ¸é¼ ¼Óµµ¸¦ ¸Å¿ì ºü¸£°Ô °³¼±ÇÏ¿´´Ù. º» ³í¹®¿¡¼´Â ±âÁ¸ ¿¬±¸ÀÇ °´Ã¼ ÀνÄÀÇ ÇÁ·¹ÀÓ¿öÅ©¸¦ ±â¹ÝÀ¸·Î ÇÏ´Â °èÃþÀû ±¸Á¶ ÆÐÅÏ ±â¹ÝÀÇ ÀÌÁø ±â¼úÀÚ »ý¼º ¹æ¹ý ¹× º¯È¿¡ °ÀÎÇÑ ÁÖ ¹æÇâÀ» ÃßÁ¤ÇÏ´Â ¹æ¹ýÀ» Á¦¾ÈÇÑ´Ù. º» ³í¹®¿¡¼ Á¦¾ÈÇÏ´Â ±â¼úÀÚÀÇ ¼º´ÉÀ» Æò°¡Çϱâ À§ÇÏ¿© ¹°Ã¼ÀÇ Á¶¸í, Å©±â, ȸÀü, ½ÃÁ¡ º¯È¯ÀÌ Æ÷ÇÔµÈ ¹°Ã¼ ÀÎ½Ä DB(KAIST-DB)¸¦ ½ÇÇè¿¡ »ç¿ëÇÑ´Ù. ½ÇÇè °á°ú Á¦¾ÈÇÏ´Â ±â¼úÀÚ´Â ½Ç½Ã°£ 󸮰¡ °¡´ÉÇÏ¸é¼ ¹°Ã¼ ÀÎ½Ä ¼º´É¿¡ ´ëÇØ ±âÁ¸ ±â¼úµé º¸´Ù ´õ ³ô°í ¾ÈÁ¤µÈ ÀνķüÀ» º¸¿´´Ù.
|
¿µ¹®³»¿ë (English Abstract) |
Due to the recent growth of mobile market, the need for descriptors that are fast and yet robust to rotation, scale and affine transformations is increasing. In this paper we propose a new binary descriptor based on layered structure patterns for object recognition and pose estimation. Recently proposed binary descriptors such as FREAK and BRISK shows similar performance as the SIFT-like descriptors and yet has greatly enhanced the processing speed. In this paper, we propose a new binary descriptor based on layered structure patterns that maintains the existing framework, and also a method to estimate the dominant orientation that is robust to variations. The proposed descriptors were tested on a database(KAIST-DB) that includes illumination, scale, rotation and affine variation. The results showed that it works in real time with better and more stable performance than recently proposed binary descriptors.
|
Å°¿öµå(Keyword) |
¹°Ã¼ÀνÄ
ÀÌÁø ±â¼úÀÚ
Ư¡ ÃßÃâ
ORB
BRISK
FREAK
object recognition
binary descriptor
feature extraction
ORB
BRISK
FREAK
|
ÆÄÀÏ÷ºÎ |
PDF ´Ù¿î·Îµå
|