| Summary: | 碩士 === 國立成功大學 === 水利及海洋工程學系專班 === 96 === This study is to recognize the usages of farms by using geographic information system with multi-temporal FORMOSAT-II satellite imagery. In the beginning combine satellite image and NDVI pixel by pixel and then process the classification results with geographic information system. Using the algorithm called spatial analyst to do zonal classifications and giving every part of farms their own attribute. The study area is the jurisdiction of the Farm Irrigation Association of Kaohsiung Taiwan. For matching the practical need, farms were classified into four group, such as rice paddy, dry farmland, fishing pool, and fallow. To recognize the usages of farms by using zonal classification method and pixel by pixel classification method in geographic information system.
The results showed using imagery in two periods of time with NDVI help to upgrade the accuracy of the recognize of farms. In the pixel by pixel case, the Overall Accuracy is 83.22% and Kappa Coefficient is 0.7633 for the result of single time period image, but by using imagery in two periods of time with NDVI help to upgrade the Overall Accuracy to 87.84% and upgrade the Kappa Coefficient to 0.8298. In the zonal case, same with using imagery in two periods of time with NDVI performed better accuracy the pixel by pixel case, the Overall Accuracy upgraded to 91.71% and the Kappa Coefficient upgraded to 0.8655.
We do not need to do the data format transform when doing the zonal classification with geographic information system, and do not have to consider the grid size and the transform methods either. The missing part cause by the process of classification is much less then traditional method. Basically, the result of the study shows it is workable to do the zonal classification and the results can be quickly presented with charts and tables by using geographic information system. The study results can be provided as a reference for the plan of replacement of the field investigation.
|
|---|
