This work is designed to exploit the current generation of high resolution panchromatic and multispectral sensors. Korean government has recently invested great amount of fund on developing earth resource satellite that will have very powerful sensor to gather a meter spatial resolution imagery. It should be developed that methods can handle the high spatial resolution satellite data. The research draws on a theoretical understanding of the spatial properties of high resolution data to develop new methods appropriate for the scale of such imagery.
This study has six chapters, including this chapter. Following the introductory material in Chapter I, Chapter II summarizes the theoretical background and previous research. The relationship between object identification and spatial resolution, and previous approaches to image segmentation for feature extraction are discussed.
The following three chapters discuss the three steps of the image analysis procedure developed in this research. Two types of data are used to develop and test the methods investigated in this research. Land parcel map as a vector data and IKONOS imagery of Jeju-shi, Jeju-Do are used to test real world applications of this research.
Chapter III describes the proposed ridge following edge detection method, and compares image segmentation based on this method to the results obtained from a square texture window. Chapter IV is an investigation of image segmentation based on the amalgamation of patches identified through ridge-following edge detection. A central aspect of this work is the development of a topology for the patches to facilitate the analysis of the spatial relationships between adjacent patches.
Chapter V proposes a region-based classification scheme using the image elements derived in the image segmentation. The approach for the classification using patch statistics is developed as a classification method using the patch mean with training data set’s probability density functions (pdfs). Finally, Chapter VI gives the overall conclusions and suggestions for using high spatial resolution remotely sensed data with this method.
This study produced a new feature-based image segmentation and classification approach specifically designed for high spatial resolution imagery. An important part of this work was the identification of image objects through the development of a reliable method of image segmentation. The new classification method resulted in improved results at both the image object scale and a richer attribution at the aggregate land cover scale.
This research made a contribution to the growing field of analysis of high spatial resolution imagery. There is a need for highly detailed information, for example in urban planning and cadastral mapping. Automated methods have the potential for more rapid and more consistent mapping than human interpretation. And the results of this research should be especially useful to various areas such as land use monitoring, urban sprawl investigation, and city and regional land use planning, etc.
- 리모트센싱을 이용한 필지별 토지이용현황 조사방법 연구(Land use/cover classification method for individual land parcel in high spatial resolution remotely sensed imagery)
- 이종열; 황승미
리모트센싱을 이용한 필지별 토지이용현황 조사방법 연구(Land use/cover classification method for individual land parcel in high spatial resolution remotely sensed imagery)
|Series Title; No||국토연 / 2002-57|
|Subject Country||South Korea(Asia and Pacific)|
|Subject||Industry and Technology < IT
Industry and Technology < Science/Technology
Territorial Development < Environment