Genetic Algorithm VS Simulated Evolution: A Comparative Study of Evolutionary Optimization Techniques for Object Recognition

One of the fundamental fields of computer vision is Object recognition and it has a plethora of applications associated with it. The primary goal of the object recognition is to identify the objects of the same type even when they are viewed from different viewpoints. However, this goal is still a very challenging research problem in the field of computer vision because of different phenomena that can modify an image such as translation, rotation, and scaling. It has been proven that shape descriptors like Fourier and Moments are invariant with respect to transformation, rotation, and scaling. However, one of the most important and challenging tasks regarding the object recognition is how to find number of descriptors of a given object. As the main objective is to maximize the recognition rate therefore another challenging question is that what is the optimum number of descriptors to be used for achieving the maximum recognition rate? Another important question is that whether all the descriptors have equal importance or not? Due to all these reasons, selection of the appropriate descriptors is of immense importance therefore applying different optimization techniques for selection of best descriptors is a key to success. In this work we do a comparative analysis of two well-known evolutionary optimization techniques known as Genetic Algorithm (GA) and Simulated Evolution (SimE). By extensive simulations in MATLAB, it is observed that Genetic Algorithm provides better performance in terms of recognition rate as compared to Simulated Evolution.