High-performance of geometric primitives detection usinig genetic algorithm

In this paper, we present some new methods for high performance of geometric primitives detection using a genetic algorithm (GA). At first, we describe the detection algorithm based on minimal subset and improvement of fitness function of geometric primitives. Secondly, we analyze the structure of minimal subsets and its probability properties in a digital image, and we improved the probability of primitive detection by reducing the invalid parts. Thirdly, we mention the subpixel measurement technique that makes edge location highly accurate, thereby increasing the accuracy of primitives by replacing the minimal subset with their subpixels. Finally, we present a method to simultaneously detect several primitives using the equivalence genes which are regarded as the set of points on a primitive; it has some excellent functions such as observation of convergence, promotion of convergence, confirmation of convergence and maintenance of multiple subpopulations.