Generic object recognition using a computer has become a necessity in various fields, such as robot vision and image retrieval in recent years. Conventional methods use conditional random field (CRF) that recognizes the class of each region using the features extracted from the local regions and the class co-occurrence between the adjoining regions. However, there is a problem with CRF – it tends to fall into the local optimal recognition result because it uses only local features and the relationship of class co-occurrence between local regions. To solve this problem, we propose a method that recognizes generic objects by incorporating bag of features (BoF) into CRF as the global feature. An experiment on an image dataset of 21 classes resulted in, the proposed method achieving an improvement in the recognition rate of 5.8%.

object recognition, conditional random fields.