Date Added: Jan 2012
Only these features which characterize the difference of similar views have value to recognize in 3D object recognition system based on view. And Principal Component Analysis (PCA) and Kernel PCA (KPCA) can extract features and reduce dimensionality. In this contribution, the authors use the PCA and KPCA to exact features firstly, and then classify the 3D objects with SVM (Support Vector Machine). In this paper, they select the Columbia Object Image Library (COIL-100) and compare the results of KPCASVM with PCA-SVM and SVM. The results clearly demonstrate that the feature extraction can greatly reduce the dimensionality of feature space without degrading the classifiers' performance.