A Study of Multiclass Object Classification using Covariance Descriptors with kernel SVM

10.29428/9789860544169.201801.0028

Chun-Yi Tsai；Wen-Hsiu Chung

Object classification ； SVM ； covariance feature ； kernel method

NCS 2017 全國計算機會議

2017（2018 / 01 / 01）

145 - 150

英文

Feature descriptor is a crucial part for image object detection and classification in computer vision. This study adopts the covariance descriptor which is a ROI(region of interest) based feature reserving the integrity of regions with rotation and scale invariant, combining with kernel SVM for multiclass object classification. The experimental results show that the combination of covariance descriptor and kernel SVM is very feasible and practical to be applied on dataset images in which the foreground objects are the major portions and the backgrounds are relatively simple.