Genetic Algorithm Based Template Optimization for a Vision System: Obstacle Detection

A simulator is developed for training and optimizing the templates for cellular neural networks for obstacle detection. The simulator uses the Genetic Algorithm (GA) for training the cellular neural network. The traditional method of genetic algorithm involves creating an initial population of random solutions (chromosomes) in binary format, the so called chromosomes encoding. But this approach of genetic algorithm defines the chromosomes in the form of real numbers, thus eliminating the need of encoding and decoding of the chromosomes. The results differ, by no means, with those of the traditional methods. The method was used for obstacle detection for autonomous vehicles giving two stereo images of a sequence as inputs. The output results for various different image processing tasks are also presented.