SHAPE CLASSIFICATION BY MOMENT AND AUTOREGRESSIVE INVARIANTS
Predrag Pejnović, MILAN MARKOVIĆ, Srđan Stanković · International Journal of Pattern Recognition and Artificial Intelligence · 2001
An experimental analysis of shape classification methods based on moment and autoregressive (AR) invariants is presented. Various types of translation, scale and rotation invariants are used to construct feature vectors for classification. The performance is evaluated using five different objects picked up from real scenes with a TV camera. Silhouettes and contours are extracted from nonoccluded two-dimensional (2D) objects rotated, scaled and translated in 3D space. The feature extraction methods are implemented and systematically tested using several parametric and nonparametric classifiers. The results clearly show the advantage of the method based on the moment invariants.