Title: On remote sensing spatial resolution: from subpixels to superpixels

题目：遥感空间分辨率：从亚像元到超像元

报告人 Dr. Xiuping Jia, The University of New South Wales

Associate Editor, IEEE Transactions on Geoscience and Remote Sensing

报告人介绍：贾秀萍博士就职于澳大利亚新南威尔士大学堪培拉分校，主要研究方向包括遥感图像处理和分类、高光谱数据分析及特征提取，是《Remote Sensing Digital Image Analysis》一书共同作者，担任多个国际期刊的编辑或者副主编，其中包括遥感信息处理著名期刊IEEE Transactions on Geoscience and Remote Sensing和IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing的副主编。发表170篇论文，其中SCI论文60余篇。

时间：9月29日（周五）上午10：00-11:30

地点：环测学院B512




A camera or a hyperspectral imager is designed with a specified spatial resolution. It is often limited, especially for hyperspectral case, where spectral measurements are the priority. The low spatial resolution leads to a large number of mixed pixels on an image, which generates high uncertainty in hard classification and inaccurate land cover monitoring in remote sensing applications. The study of spatial-resolution improvement via image processing techniques, including superresolution reconstruction and spectral unmixing, has been conducted actively in remote sensing data analysis, which offers an effective means to overcome the hardware limitation.



On the other hand, a given resolution, which is low for one application, can be high for another application. The high spatial resolution increases intra class variability and makes object recognition difficult. To address these problems, superpixel based image classification techniques have been developed in recent years.



In this talk, the concept of pixel, subpixel and superpixel will be introduced. Subpixel and superpixel based image classification techniques will be overviewed and discussed.