Association Rules Mining Techniques Used For Remote Sensing Data And Applications

Data mining is a process which extracting potential and valuable models or rules from large data sets. Since 1970's, long-range accumulation of remote sensing data and its productions had been achieved with the rapid development of collecting technique for remote sensing data. The fact that processing technique lags behind collecting technique of remote sensing data can not meet national "10th five-years plan" important demands, for example resources survey, environment, nature disasters preventing and assessment, carbon circulating, map spectrum. So that data mining techniques were listed in national 863 information, processing and application field in 2002. This dissertation introduced some research achievements in association rule mining in corresponding to the research works in doctoral period. As we observed in gee-phenomena themes and may frequently meet problems such as, multi-dimensional correlation, temporal sequential correlation and feature correlation. For interests and application, the author selected association rules of multi-dimensional association, temporal sequential association, and feature association as study focal points. The better results are arrived in terms of application and the scope to look at feature extraction.(1) Multi-dimensional association rule mininghi geo-science field, one phenomenon is occurred, there are at least two factors in action. This kind issue can be induced into multi-factor associations. Finding rules or models among multi-dimensional associations can regard as question of multi-dimensional association rule mining, hi the dissertation, an example is given to show multi-dimensional association mining used in project of "closing farm land for forest" in MinJiang river drainage in SiChuan province. It discussed rules of soil erosion which resulted from integrated affects about slope, vegetation cover and farmland, established multi-dimensional correlative rules. Slope decision-makingfarmland, established multi-dimensional correlative rules. Slope decision-making rules and minimal vegetation cover rules of "closing farm land for forest" achieved with multi-dimensional association rules mining will be useful to make decision of management.(2) Temporal sequential association rule miningWhen a geo-phenomenon is observed with time sequential changes, some association rules can be revealed between main processing and background associate factors. The dissertation took dust storm as an example to show temporal sequential association rule mining procedure. In dust weather procedure in spring, 2002, LCT (land surface temperature) and SM (soil moisture) can be retrieved from NOAA/AVHRR CH3, 4, 5, by sampling three months data starting from before dust storm to end of dust storm. The result showed there relationships are opened out inherent dry dust area and retrieved temperature and moisture image data.(3) Feature association rule miningNormally, features of objects occur with features of background, so that many problems of extracting information from remote sensing data can be generalized to features of object and background. Feature association rule mining is filtering and finding characteristic patterns between object and background. In the dissertation, an example of extracting remote sensing canopy spectrum of vegetation and its mechanism connections with geo-chemistry trace elements Co-Mn-B-Mo-Zn was given.In conclusion, only when the data mining used in the process of remote sensing data collecting, processing and application, the scope of processing and extracting is focused on rules and models, the nature of phenomenon and the mechanism of information can be revealed and recognized. From the study, it is proved that data mining technique has great potential roles in dealing with large amount of remote sensing database. Association rule mining is an important part of data mining. The three aspects of association rule mining of remote sensing data presented in the dissertation some samples of application of them. There are lots of aspects els...