Reclamation Information Extraction Using Decision Tree Method In Jinzhou Bay Area

In this paper,the key elements and classification system of land reclamations were established on the multi-scene remote sensed image,such as:Landsat7ETM+and Landsat8OIL data,by choosing Jinzhou Bay as the study area,where the data were used in dates respectively,2000,2005,2010,2015.The classification method based on decision tree was selected.The types of reclaimed land in 2015 Landsat8OIL images were classified.The accuracy of the decision tree method is verified by comparing the precision with the supervised classification method,and the method with high classification accuracy is selected to deal with the remaining image by the same method.Statistics and change information of reclamation types in the study area for more than ten years are obtained.The results of this paper are as follows:(1)The definition of sea area use types and the actual situation of the study area based on the Classification system of Sea area use and the Code of Marine Survey and the practical problems in remote sensing interpretation,The classification system and remote sensing interpretation mark of Jinzhou Bay sea area were established.By using two-level differential classification system,the reclaimed land and unused land in the study area were divided into the first class,and the reclaimed sea area was divided into salt industry sea and culture sea.The reclaimed land is divided into three classes:construction reclamation,agricultural reclamation and unconstructed land.The unused land includes two classes:unused water area and tidal flat.(2)Extracting the information of the type of reclamation in 2015 Landsat8OIL remote sensing image,and using the spectral information,texture information and shape information of the image to establish the decision tree classification model.Comparing the result of information extraction with that of using supervised method,the total classification accuracy of decision tree method is 90.75%,and the total precision of supervised classification method is 80.50%.The former is obviously more accurate than the latter.(3)Using the decision tree classification method to classify and extract the type information of the remaining images,and analyze the classification results of the multi-period images,and draw the conclusion that the area of unused land is gradually decreasing in the study area.From 2000 482.5km~2 to 2015 year 114.92 km ~2.Reclamation area increased from 2000 year 88.82km~2 to 215.98 km ~2.Reclamation area increased from 2000year 10.1km~2 to 2015 year 144.92 km~2.