Study on Multi-scale Segmentation of High Resolution Aerial and Satellite Images
JIANG Miao, CHEN Xiuwan, WU Caicong et al.

北京大学遥感与地理信息系统研究所，北京 100871

Abstract

This article analyses the aerial image, QuickBird and IKONOS images which cover the same area in the view of multi-scale segmentation，then discusses the different segmental effects under different segmental scales in different images. To boost up the recognizing ability, and to advance the precision and efficiency of the applications with aerial image and satellite image data, we discuss the best segmental scale to different objects and different images, and compare the final classification result. The result indicates the best segmental scale for aerial image, QuickBird and IKONOS image is 125,100 and 75 respectively. The classification precision of QuickBird is the best. Aerial image and IKONOS image’s classification precision is approximate. So in the practical application, we can use aerial image to replace high resolution satellite images.
Keywords:  multi-scale segmentation; high resolution; aerial image; best segmental scale