講座題目：非線性高光譜解混的新進展 New Developments in Nonlinear Hyperspectral Unmixing

報 告 人：Paul Scheunders

時 　 間：2019年10月25日(周五)10:00-12:00

地 　 點：中關村校區信息實驗樓202報告廳

主辦單位：意昂平台、信息與電子學院

報名方式：登錄意昂官网微信企業號---第二課堂---課程報名中選擇“【百家大講堂】第252期：非線性高光譜解混的新進展”

【主講人簡介】

  1990年，Paul Scheunders在比利時安特衛普大學獲得了統計力學領域的物理學博士學位🧙‍♂️。1992年，他成為安特衛普大學物理系視覺實驗室的一名助理研究員，目前是該實驗室的一名教授🧱👶🏻。他目前的研究興趣是遙感圖像處理，尤其是高光譜圖像處理。他在圖像處理、模式識別和遙感領域的國際期刊和會議記錄上發表了250多篇論文。
  Paul Scheunders是IEEE Transactions on Geoscience and Remote Sensing 和Remote Sensing (MDPI)的副主編，並在許多國際會議上擔任項目委員會成員😋，還是IEEE Geoscience and Remote Sensing Society的高級會員。

Paul Scheunders received the Ph.D. degree in physics, with work in the field of statistical mechanics, from the University of Antwerp, Antwerp, Belgium, in 1990. In 1992, he became a research associate with the Vision Lab, Department of Physics, University of Antwerp, where he is currently a professor. His current research interest includes remote sensing and in particular hyperspectral image processing. He has published over 250 papers in international journals and conference proceedings in the field of image processing, pattern recognition and remote sensing.

Paul Scheunders is Associate Editor of the IEEE Transactions on Geoscience and Remote Sensing and of Remote Sensing (MDPI) and has served as program committee member in numerous international conferences. He is senior member of the IEEE Geoscience and Remote Sensing Society.

【講座信息】

  光譜解混是以信號的純組分（端元光譜）和端元豐度的函數描述高光譜信號的過程。由於植被的多重反射或精細混合以及混合和復合材料化學性質的變化，光譜反射率呈現高度非線性👨🏻‍🔬。本次報告首先通過光譜解混過程的圖示來討論一種基於模型的非線性光譜解混方法。將重點討論一種多線性解混模型以及包含陰影效應的拓展🧘‍♂️。其次，報告將討論一種新的采用數據驅動的非線性光譜解混，其需要真值的端元光譜和豐度。該方法適用於很多近距離應用，如巖芯樣品📴、混合和復合材料的表征以及葉片參數估計。

Spectral unmixing is the process of describing a hyperspectral signal in function of its pure constituents (endmember spectra) and their fractional abundances. The spectral reflectance can show highly nonlinear behavior, because of multiple reflections in vegetation or intimate mixing and changes in chemical properties of mixed and compound materials. This talk will first discuss a model-based approach for nonlinear spectral unmixing based on a graph description of the spectral mixing process. In particular, a multilinear mixing model and an extension including shadow effects will be discussed. Secondly, the talk will discuss a new data-driven approach to nonlinear spectral unmixing, which requires ground truth endmember spectra and fractional abundances. This approach will be applied to a number of close range applications such as the characterization of drill core samples, mixed and compound materials, as well as leaf parameter estimation.