Hybrid genetic optimization and statistical model-based approach for the classification of shadow shapes in sonar imagery

We present an original statistical classification method using a deformable template model to separate natural objects from man-made objects in an ima ge provided by a high resolution sonar. A prior knowledge of the manufactur ed object shadow shape is captured by a prototype template, along with a se t of admissible linear transformations, to take into account the shape vari ability. Then, the classification problem is defined as a two-step process. First, the detection problem of a region of interest in the input image is stated as the minimization of a cost function. Second, the value of this f unction at convergence allows one to determine whether the desired object i s present or not in the sonar image. The energy minimization problem is tac kled using relaxation techniques. In this context, we compare the results o btained with a deterministic relaxation technique (a gradient-based algorit hm) and two stochastic relaxation methods: Simulated Annealing (SA) and a h ybrid Genetic Algorithm (GA). This latter method has been successfully test ed on real and synthetic sonar images, yielding very promising results.