3维物体SIFT特征的提取与应用_熊英37.pdfVIP

15 5 Vol. 15, No. 5 2010 5 Journal of Image and G raphics M ay, 2010 3 SIFT 熊 英 马惠敏 (, 100084) SIFT ( scale-invariant feature transform ), ( ), ( SIFT ) , , 3 SIFT 3, , SIFT , SIFT , , SIFT 3 , , SIFT 3 : TP391. 41 : A : 100 -89 1( 2010) 05-814-0 Extraction and Application of 3D Object SIFT Feature XIONG Y ing, MA Hu mi in ( In stitu te of Imag e and Grap hics, D ep artm ent of E lectronic Engineering, T singhua Un iversity, B eij ing 100084) Abstract The S IFT algorithm isw idely adopted by researchers in mi age and graphic study, w ith its m any advan tages such as invariance to scaling, rotation, noise and illum ination changes. SIFT feature is based on local grad ient, m aking it invulnerable to large scale of mi age extension, com pression and rotation, and th is m eets the practical requ irem ents of 3D object recognition. And the sensitivity to hom ographic transform ation of the feature can be applied to partition the view space. Reasonab ly setting the th reshold value, SIFT algorithm can handle the techn ical problem s such as cutting the object from its background. A fter pre-processing, the h igh computation comp lexity can be reduced, mak ing the system run in real tmi e. Therefore, app lying S IFT feature in view space partition, cutting the object from its background and pattern m atch ing can effectively enhance the robustness of the system and mi prove its speed and efficiency. K eywords SIFT, 3D object recognition, pattern m atch, view space partition 0 ,