| Remote sensing technology has the characteristics of magnanimity data. Variouskinds of satellites to ground observation can constantly offer remote sensing images ofdifferent space resolution ratio, spectrum resolution ratio and time resolution ratio. Thesingle information source is difficult to fully meet the need of the resources monitoring.In order to fully utilize the data source, find the needed information from themagnanimity data and have a more overall, clearer and accurately understanding andcognition to the observing objects, people imminently wish to seek a technologicalmethod which can comprehensively utilize the data. Data fusion offers a feasibledirection on comprehensive utilization of RS data and the improvement of data quality.It has been a focus in the international remote sensing field in recent years all the time. As a comprehensive technology of processing data, remote sensing data fusioninvolves a wide field and many contents. It is a complicated technology and can belimited by a lot of factors. It's final purpose is to choose suitable fusion algorithm formeeting precision demand of specific application. There are no unanimous appraisalcriterion and standard about the result of data fusion at present. So, the objectivity andcredibility of data fusion appraising is the question that people have been exploring allthe time. This paper plans to do the research on the appraisal of fusion result and discussthe fusion methods on the basis of something studied from the forefathers. First of all,the results were analyzed to evaluate five fusion models on spectral integrity andspatial characteristic qualitatively and quantitatively. Then, it takes the comparisonthrough supervised and unsupervised classification result for the case district to provedifferent influence produced toward different RS fusion models. It can offer somereference suggestions for the analysts of different applications. This paper is divided into six parts including preface and conclusion.In the first part,has introduced the research meaning of the paper from thepeople's need to the remote sensing data and the complexities of data fusiontechnology and analyzed the characteristic of the image data and the case district . In the second part, has made a general survey of the domestic and internationalremote sensing data fusion research. It has described the RS data fusion theories, keydata fusion technique, the application of fusion technology and the existing problems. In the third part, has introduced the characteristics of the remote sensing and thedata fusion algorithms. It has analyzed the SPOT5 image data, data preparation ofimage and had an analysis of bands combination. Then it describes the researchmethod of this paper. In the forth part, has analyzed to evaluate each of the five basic fusion modelsqualitatively and quantitatively. It has analyzed from the aspects of basic statisticscharacteristics, relevant characteristic and spectrum curve of the surface featurespectrum, etc, to evaluate their spectral integrity and spatial characteristics. Theanalysis and results can offer the foundation for the perfection of the fusion algorithm. In the fifth part, has taken almost the same test procedure to classify the differentdata sets and then compared the change of area of all kinds of surface feature in theclassified image to prove the influence data fusion models has on RS imageclassification. Through the analyzed result, it has had further analysis on theclassification accuracy for the original multispectral image and the fusion imagerieswhich can be used as the foundational image for supervised classification andunsupervised classification. In the sixth part, has summarized the results and revelation of research and theexisting problems... |
