Japanese / English


論文の言語 日本語
著者 Katsufumi Inoue,Hiroshi Miyake,Koichi Kise
論文名 A Memory Reduction Method for 3D Object Recognition Based on Selection of Local Features
論文誌名 Proceedings of Third Korea-Japan Joint Workshop on Pattern Recognition (KJPR2008)
ページ pp.7-8
年月 2008年11月
要約 3D object recognition methods based on local features employ local features extracted from a large number of images for constructing object models and recognize 3D objects with these models. An important problem of these methods is requirement of huge memory space due to a large number of local features that constitute the object models. In this paper, we propose a simple method to reduce the required memory space by selecting local features in the object models. The proposed method is characterized by the mechanism of selection which employs estimates of the positive and negative effects of each local feature for object recognition. From experimental results for 30 objects, we achieved the recognition rate of 98.8% with about 10% of local features extracted from all images for models. Experimental results with COIL-100 (Columbia Object Image Library-100) show that 1/6 of the total local features allowed us the recognition rage of 96.8%.