A novel object recognition system for cognitive robot

Cognitive robot is so amazing and challengeable for its human-like behavior, especially for object recognition which goes through detection and recognition processes. Each part will be involved with the difficult technologies. In this paper a human-like object recognition system is built and three hypothesizes are proposed which based on cognitive psychology. This system aims to address objects which refers to detect, to recognize and to learn. In learning part, it could be trained based on learning algorithms. New category or new ascription will be added in this cognitive system through human-robot interaction. Additionally, three experiments were implemented to show the validity of our system. Target object is located from indoor complicate scene in the first experiment. Then, the second experiment aims to recognize given object through searching and matching from given image library. The third experiment shows how to find a special object with given class which performs very well in indoor scenario.