An Approach of Reservoirs and Ponds Extraction Based on Pixels and Object-Oriented Method

Multi-scale segmentation occupies lots of system resources and time.Traditional object-oriented classification method is based on the panoramic image and dismembered objects,the results of which need to be improved artificially.This paper proposes a method based on pixels combined with object-oriented classification to extract information in high-resolution image to solve these problems.A rough classification is performed by using some bands of the original QuickBird satellite imagery to create a sub-project of low-resolution image.Then an object region is analyzed in detail one by one.The contours of extracted objects are normalized thereafter.Finally,the extraction of reservoirs and ponds is taken as an example.Experimental results using a wide range of image show that this method improves the efficiency of information extraction with a fine overall effect.