An Adaptive Object Perception System Based on Environment Exploration and Bayesian Learning

Cognitive robotics looks at human cognition as a source of inspiration for automatic perception capabilities that will allow robots to learn and reason out how to behave in response to complex goals. For instance, humans learn to recognize object categories ceaselessly over time. This ability to refine knowledge from the set of accumulated experiences facilitates the adaptation to new environments. Inspired by such abilities, this paper proposes an efficient approach towards 3D object category learning and recognition in an interactive and open-ended manner. To achieve this goal, this paper focuses on two state-of-the-art questions: (i) How to use unsupervised object exploration to construct a dictionary of visual words for representing objects in a highly compact and distinctive way. (ii) How to learn incrementally probabilistic models of object categories to achieve adaptability. To examine the performance of the proposed approach, a quantitative evaluation and a qualitative analysis are used. The experimental results showed the fulfilling performance of this approach on different types of objects. The proposed system is able to interact with human users and learn new object categories over time.