Multiple Source Remote Sensing Image Fusion Technology And Applied Research

With the rapid development of modern remote sensing technologies, acquiring information from remote sensing image has become an important ways to acquire information, because remote sensing technologies possess macro, dynamic, rapid, exact and integrative advantages.In actual application of remote sensing, remote sensing data have different resolution characteristic of space, spectrum and time. Fusion technologies of remote sensing image have become an important tool for analyzing remote sensing image, because fusion technologies of remote sensing image can improve resolution characteristic of space, enhance features of object, improve precision of classification and dynamic detection as well as complementary capability of information. This dissertation studies outstanding problems in remote sensing image fusion, such as: (a) preprocessing technologies of image fusion; (b) robust fusion algorithms, new fusion models, as well as effective evaluation of fusion results; (c) application technologies of fusion. The overall research embodies current difficulties and emphases in remote sensing image fusion.The main contributions of this dissertation are as follows:(1) Preprocessing algorithms:(a) After analyzing characteristics of noise of remote sensing image, a new algorithm for detecting and eliminating thin linear noise, salt and pepper noise is proposed. Simulation results show that the algorithm has advantages over those general algorithms (such as medium filters, state statistical filters, Wiener filters) and latest algorithms (such as self-adaptive fuzzy multi-level medium filters, and algorithm for eliminating noise based on neighborhood information) about removing noise and protecting detail and structure information of image.(b) Considering the difficulties of extracting matched structures in image registration algorithms, a new method for detecting and matching control points is proposed, which can be realized automatically and rapidly. The method improves velocity and exactness greatly compared with current algorithms looking for matched points, therefore it provides for convenience for using better matching model (such as triangle network model).(c) Designed self-adaptive filter can not only eliminate point noise effectively induced by image processing, but also pick up interesting objects with certain size.All algorithms offer basis for optimizing fusion results.(2) On the basis of theory and practice of image fusion evaluation, general principles for fusion evaluation are induced, on the other hand, a method...