الفهرس 
ACKNOWLEDGMENTSOnly 14 pages are availabe for public view
 |  | from | 106 |  |  |
| | | from | 106 | | |
Abstract
Digital elevation model (DEM) is remote sensing technique that using to represent or reconstruction topography model of the earth’s surface. Recently the need to use the digital elevation model has increased in numerous applications in earth environmental especially assessment water resources. Error is the main problem in DEM which produced by multiple sources like incomplete density of observations, spatial sampling and processing errors. These errors in finally consist of uncertainties region or uncertainties pixel in digital elevation models. In this work the Shuttle Radar Topography Mission digital elevation model and global position system survey points were used to visualize the height differences as 2- dimension raster image using Inverse Distance Weighted interpolation in Geographic Information System. Then we used deep learning model to predict uncertainties regions based on training differences image and shape files that corresponding features of errors.
Water catchment area is one of important water resources that can produce many benefits. Then the water catchment region had been detected using powerful depression algorithm (level set algorithm) and DenseNet model. The level set algorithm is based on the real sinks, which are represented by a binary image mask. The level set algorithm used DEM and image mask as input data. Polygon features representing catchment areas produced in shape files are the result of this method. These files used as input data to DenseNet model to detect catchments by training data. The DenseNet model has demonstrated its effectiveness in recognizing catchment regions with accuracy 92.7% from free error DEM.