Remote Sensing Rangeland Detection using Geographic Object Based Image Analysis and Data Mining

Remote Sensing Rangeland Detection using Geographic Object Based Image Analysis and Data Mining

Rangelands are the most widespread type of Land Cover. The main goal of this research was to develop a procedure to classify rangelands in the remote sensing images. To this end, two important techniques were joined: Geographic Object Based Image Analysis and Data Mining. J44 data mining algorithm was used to create the required knowledge for creating rule set and GOBIA was used to classify the image. Study area is located in the north-east of Iran in upstream of Gorganrood watershed and is the start point of several floods. A Landsat 4 image was selected for the rangeland classification. After the segmentation selected samples were imported to data mining software and afterward generated knowledge was applied on the image. Classification accuracy was calculated using 121 reference point and overall accuracy of the created map and Kappa were 59.5 and 59.3, respectively. Ultimately, it should be mention that using combined GOBIA and data mining in rangeland mapping can be effective and beneficial.