Biologically Inspired Model with Feature Selection for Target Recognition Using Biogeography-Based Optimization

To detect salient ground targets precisely and rapidly during aerial reconnaissance, this paper describes a novel object recognition method based on the feature selection of a biologically inspired model and biogeography-based optimization. As a promising approach to object recognition, the biologically inspired model is a hierarchical system of building an increasingly complex and invariant feature representation, which closely follows the process of object recognition in the visual cortex. These scale- and position-tolerant features are constructed by alternating between a template-matching and a maximum-pooling operation. Because of the many patches extracted in the standard biologically inspired model, the random mechanism may extract patches from irrelevant parts of an image and consume a lot of time. In this work, a feature selection method is proposed based on a new population-based evolutionary algorithm called biogeography-based optimization to choose the proper set of patches with high accuracy of classification and recognition. A support vector machine classifier is used for evaluation of the fitness function in biogeography-based optimization and to calculate the recognition rate in testing. A series of experiments are conducted, and the comparative results demonstrate the feasibility and effectiveness of the approach.