Collaborative hybrid classifier learning with ant colony optimization in wireless multimedia sensor networks

Wireless multimedia sensor network (WMSN) has powerful multimedia signal acquisition and processing abilities. This paper proposes a collaborative hybrid classifier learning algorithm to achieve online support vector machine (SVM) learning for robust target classification in WMSN. The proposed algorithm is carried out in a hybrid computing paradigm, which combines the advantages of progressive computing paradigm and P2P computing paradigm. Importantly, the participant sensor nodes are purposefully selected by evaluating the specific effectiveness. With the sensor nodes selection strategy, the energy consumption and the impact of inevitable missing detection and false detection can be reduced. Besides, ant colony optimization is also used for decreasing the energy consumption in routing. Experimental results demonstrate that the collaborative hybrid classifier learning algorithm can effectively implement target classification in WMSN, and the ant colony optimization based routing and clustering method can largely decrease the energy consumption and time cost.