The Study On Genetic Optimization And Related Combination Algorithm For Remote Sensing Data Processing

Nowadays, series of modern plans by using satellites for remote observation of the earth provide us scientific platforms and new opportunities. At the same time, modern computer technology provides powerful computing and data processing abilities. And varies new algorithms of data processing, modeling and prediction provide useful tools for further mining the information from large amount of satellite data. At present, many researches are still trying to find new efficient algorithms for remote sensing data processing from different approaches, and hoping those algorithms have abilities of self-understanding, self-recognizing, self-learning and self-adapting. To fully mine all information efficiently for the remote sensing data, which includes spectral, spatial and angle information, etc, is the objective and trend of new researches.In this paper, we focus on applying modern intelligent tools such as Genetic algorithm and Bayesian network to remote sensing data processing. Genetic algorithm elucidates the principle to understand the world in the evolutional way and provides an optimization method for data and information processing. The advantage of Genetic algorithm includes global searching and parallel computing, which also makes it popular in many other fields. Bayesian network is a new pattern classification graph model based on Bayesian statistics, it is an intelligent tool that can integrate prior knowledge and sample information in classification and causality discrimination for data processing. Thus, it is hot and in the research frontier to apply Genetic algorithm and Bayesian network to the remote sensing data processing. In this paper, Genetic algorithm is presented as the mainstream, combined with Bayesiannetwork, a new method for remote sensing data processing is proposed from views of image matching and pattern classification. There are totally seven chapters, which are listed as the following:In the first chapter, recent researches in both Chinese and international literatures are surveyed on applying Genetic algorithm to remote sensing; also the whole architecture and main techniques in this paper are roughly presented. The second chapter introduces the theory of Genetic algorithm, including the mathematical principle, evolutional directions, evolutional rules and main types. In the third chapter, based on the previous works of other groups, the hyper-plane classification model is presented, an algorithm of genetic hyper-plane classification is fully discussed; the algorithm is also evaluated from several aspects such as the simple classification, control parameter setting, whole classification efficiency by using data from MODIS, ETM, ASTER, etc. In the fourth chapter, by studying aerial images of a certain area in Beijing, the Genetic algorithm is applied for remote sensing image matching, which are supported by experiments. In the fifth chapter, Genetic algorithm is combined with Bayesian network for image classification and prediction for the remote sensing data. Based on the learning algorithm of Bayesian network classifiers, the genetic learning process is introduced in details. In the sixth chapter, by combing Genetic algorithm into image segmentation, a new frame of genetic based quad-tree classification algorithm is proposed. In this chapter, image segmentation and edge detection is first introduced, then the quad-tree representation of remote sensing image is fully studied. Comparing to other algorithms, the advantage of this new strategy is that it can use spatial and spectral information at the same time, which is a developing trend of remote sensing data processing. In the seventh chapter, conclusions are made and discussions for further researches are proposed.The main innovations and values of this paper are listed as following: 1, Further study Genetic hyper-plane classification algorithm for remote sensing data processing, and carefully compare binary encode system and decimal encode system in this algorithm, which gives a new way for data pattern representation andIV...