Researches On Genetic Algorithms And Its Applications In Tracking System

Genetic algorithms(GAs) have emerged as practical, robust optimization and search methods in the last years. Diverse areas such as function optimization, computer vision, engineering structural optimization, machine learning, etc. have profited from these methods. Researchers also made progress in study on running mechanism and performance analysis of GAs. In the dissertation, researches have focused on analyzing and improving GAs performance, and introducing GAs into tracking system. The emergence of new GAs implementations for better performance has been accompanied by considerable theoretical research, especially in developing models of GAs dynamics, analyzing problems that are hard for GAs, and, most important, gaining a deeper understanding of how GAs work. Afler having studied how GAs work, a new function for quantitative analysis of GAs performance is proposed, which is more suitable for practical optimization problem. An efficient approach is presented for adaptive adjusting the crossover and mutation probabilities of GAs as improving the performance. In order to overcome drawbacks of tradition thresholding methods, a minimum error segmentation method and a multi-parameter GAs by using the motion information of objects based on genetic algorithms are presented. Since different segmentation algorithms works well for different applications, quality evaluation of image segmentation is indispensable, and thus we introduce fuzzy measure into quality evaluation algorithm of image segmentation and propose a function that could change an image to the field of fuzzy property effectively. There are several goodness measures available, such as intra-region uniformity, contrast, and region shape. But sometime, only one of these goodness measures can not judge the quality of segmentation results properly, so a composite evaluation function that includes region uniformity, gray contrast, shape parameter and fuzzy measure is presented. Finally we draw GAs into tracking system, develop a fast correlation matching algorithm and an object detection and localization method based on GAs. Moreover, we propose a scheme for implementing a tracking process using GAs.