Two-stage neural network for recognizing partially occluded objects

Abstract

This paper proposes a two-stage method for recognizing partially occluded objects. In the first stage, the features of the single objects are used for the back-propagation neural network (BPNN) to classify the features of the occluded objects. When the features have been classified, those belonging to the same class are then matched to single objects. In the second stage, the matched features of the single objects are considered as the test data in the BPNN. The probabilities of each single object appearing in the occluded object are found. The experimental results indicate that the proposed method can recognize occluded objects effectively.

Keywords:

• MACHINE VISION
• FEATURE EXTRACTION
• FEATURE CLUSTERING
• BACK-PROPAGATION NEURAL NETWORK
• OCCLUDED OBJECT RECOGNITION