Study On Fusion Techniques Of Multi-Sensor Remote Sensing Data And Extraction Of Desertification Information

With the development of modern remote sensing technology,more and more multi-resolution,multi-temporal and multi-spectral remote- sensing image data have been receiving by different kinds of remote sensors which had been launched successfully one after another. It becomes increasingly the demand and trend to processing the multi-source remote sensing images which were combined in order to improve the reliability of visual judging, the capability of interpretation and veracity of data classfication and targets recognition.The fusion technology of the multi-source remote sensing data was the appropriate algorithm which integrated organicaly with the different images of same area by spatial registration and obtained complementary information from each other to meet different demand. According to different levels of information attribute, there were three types of the multi-source remote sensing data fusional technic,namely, the pixel level, characteristic level and decision-making level.Land desertification is one of the most important issues all over the world. It is an important breakthrough for remote sensing technology to position the area in order to control the issue effectively.This paper is aimed to reasearch on multi-source images fusion technology system,namely to find out a set of the more effective algorithm to improve the classification precision of land desertification and extract desertification information exactly.This paper included three main contents such as:1.The first part involved in the research that had used the pixal-level fusional algorithms such as PCA,Mutiplicative,Brovey,Histogram matching HIS,Wavelet based on PCA,Wavelet based on HIS to experiment,and then choose the best fusional algorithm to improve classification precision.The result showed that the PCA algorithm most fitted to extract desertification information,which the precision enhanced 8.17%.2.The second phase was to experiment with feature-level fusion algorithms. The experiment chose four texture feature indices such as variance entropy homogeneity and correlation to analyze. The result showed that the homogeneity index was the most helpful character to improve the classification precision of land desertification which heightened 1.81%;3.The third content consisted of decision-level fusion technology and classification algorithm. The paper summarized present literatures about decision fusion algorithm to find that the C4.5 algorithm had the advantages which could change the continuous variable into discrete variable ,and process the training samples with lacking some features ,and take order with imperfection data.Finally, the C4.5 algorithm had been taken up to finish a course of action based on PCA and the texture features, for example established the decision-tree and set up the classification rules and trained the model step by step to achieve ultimate object of endeavor, and the classification precision increased 17.4%.It made up the method system to improve the land desertification classification precision,and named as a new fusion technology based on PCA and image texture characters.There were two innovations in this paper:1.The study found out a technique system to improve the classification precision of land desertification.2.This paper advanced a new technique-----a Fusion based on PCA and image texture characters.It could result in different conclusion to adopt other sensor images,on the other hand to resolve the other study object and task.So that ,this issue awaited following study.