Research On Multi - Scale Segmentation Technology In High Resolution Remote Sensing Image

With the rapid development in remote sensing technology, it has been more and more applications to the measurement, land, agriculture, transportation, forestry, water conservancy, military and other industries and fields, especially the emergence of high-resolution remote sensing in recent years, from it we can get more and more geographic information as data for the future development of surveying and mapping, and it provides a new direction of remote sensing science.High-resolution remote sensing has huge advantages in terms of resolution,at the same time provides a clear shape,structure and so on information about geometry. However, we use the traditional method based on image elements extract feature form high-resolution remote sensing image, not only its accuracy and efficiency are not good, but also wastes a lot of spatial information.This article chooses two representative WorldView-2 images about Baiyin City, Gansu Province,using Multi-scale segmentation method based eCognition software,in other word,using a variety of scale on the images to segment,and it analyses and concludes segmentation results.while flexible use of the feature itself spectral information and geometry information, thematic layers can be imported and directly applied to the segmentation and classification,,with loading people thinking, well-written set of rules,finally different scales of geographic information can be extracted by different of layers and classes.In the last article,it compares the results of Object-oriented method with traditional manual interpretation method,the former not only take full advantage of large quantity spatial from information high-resolution remote sensing images, and have faster process, more accurate extraction.