Efficient Resource Allocation using a Multiobjective Utility Optimisation Method

Abstract : In this paper we present an extension for two recent active vision systems proposed in Navalpakkam and Itti [1], and in Frintrop [2]. The novelty of our proposed system is twofold: first it extends the existing approaches using both prior and dynamic contextual knowledge, enabling to adapt the proposed system to the present environment. Second, the decision making process intuitively used in [1, 2] is formalised in this paper and put into the context of multiobjective optimisation using a utility concept. We discuss three different saliency algorithms to be used in the system as well as three different methods to determine common utility. Our presented system is quantitatively evaluated using a motorway traffic sequence recorded by a test vehicle equipped with a multimodal sensor system.