Doctor的 Defense_final.pptxVIP

Application of Spatial Features to Classification, Segmentation, and Sharpening in Remotely Sensed Images CHEN, Xuehong Supervisor: Prof. Yasushi Yamaguchi Graduate School of Environmental Studies, Nagoya University Aug., 2012 1 Outline 2 Introduction 3 1.1 Remote Sensing Remote sensing is the science (to some extent, art) of obtaining and analyzing the information of the earth surface by electromagnetic wave without making physical contact (Lillesand and Kiefer, 2000). There are three dimensions of information in the remotely sensed data. Temporal Dimension 4 Spatial Feature 1.2 What is Spatial Feature? There is no strict definition (Materka Strzelecki, 1998). Spatial distribution of geographic elements are NOT independent. Spatial features include: Spatial Autocorrelation … 5 Land cover classification Label each pixel in a scene to specific land cover types. Image Segmentation Subdivide an image into a set of “homogeneous” objects. Image Sharpening Resample a low resolution image to a high resolution image by fusing other information source. 6 1.3 Classification, Segmentation, and Sharpening 1.4 How to Describe Spatial Features? Moving-Window Based Method Spatial Domain Filter, Markov Random Field, Mathematical Morphology, Co-occurrence Matrix Object Based Method Image Segmentation Image Based Method Semivariogram, Fourier Transformation 7 1.5 Problems in the Previous Studies Problems: Particular geographic characteristics were seldom considered. Applications were mostly conducted on classification or change detection. 8 Moving Window Based Object Based Image Based Land cover classification; Change detection; Interest object detection; Endmember extraction (e.g. Bruzzone Prieto, 2000; Plaza et al., 2002) Scale effect of gray image; Stripe noise removal (e.g. Chen et al., 2004) Methods Previous Applications My study aims to: Consider more on geographic characteristics Extends the application of spatial features to more research fields 1.6 Framework Mo