海外名师大讲堂 — On hard and soft hyperspectral image classification uncertainty

讲座名称: 海外名师大讲堂 — On hard and soft hyperspectral image classification uncertainty
讲座时间: 2018-06-21
讲座人: Xiuping Jia
形式:
校区: 兴庆校区
实践学分:
讲座内容: 海外名师大讲堂 — On hard and soft hyperspectral image classification uncertainty 报告时间: 2018年6月21日(周四) 下午14:30 报告地点: 仲英楼(理学院)B-253 报 告 人:Xiuping Jia 报告摘要: A hyperspectral imager is designed with a specified spatial resolution, which is often limited as spectral measurements are the priority. The low spatial resolution leads to a large number of mixed pixels in an image, which generates high uncertainty in hard classification, where each pixel in a hyperspectral image is labelled as one of the defined classes. Spectral unmixing techniques have been developed to map each pixel as a weighted sum of the endmember classes it contains. It offers an effective means to overcome the hardware limitation via post image processing. Several issues in spectral unmixing techniques will be discussed in this talk, including endmember extraction, primary class data representation, and mixture modelling. The emphasis will be placed on within class spectral variation and the associated unmixing uncertainty. The concept of unmixing resolution and fuzzy assessment will be introduced and discussed.
相关视频