通知通告

GIS学术讲座 Automated Terrain Feature Identification from Remote Sensing Imagery: A Deep Learning Approach

文章来源:  |  发布时间:2018-05-28  |  【打印】 【关闭

  

  报告题目:Automated Terrain Feature Identification from Remote Sensing Imagery: A Deep Learning Approach 

  主 讲 人:李文雯 副教授(美国亚利桑那州立大学地理科学与城市规划学院) 

    

  报告摘要:This talk introduces a novel deep learning framework to support terrain feature identification for automated landscape scene recognition and interpretation. Developing landscape scene recognition techniques are of critical importance for geospatial researchers to differentiate terrains, support autonomous driving and navigation, detect interesting targets, identify anomalies, aid search and rescue operations, inspect living conditions of refugee camps, among others. This work takes advantage of the rapid advancement in deep learning (DL) and convolutional neural networks (CNN), the maturation of high-performance computing hardware chips, and cloud platform as well as the proliferation of geospatial big data. It achieves automatic and high-performance terrain feature detection in the following ways: (1) it extends a popular object detection network by integrating deep and deeper CNNs, boosting up its capability in feature extraction; (2) it provides a comprehensive evaluation to identify the best combination of hyper-parameters that optimize the model performance; (3) it introduces ensemble techniques to further enhance the model’s predictive accuracy by the collective intelligence of multiple classifiers. We expect this work to contribute significantly to the advancement of terrain analysis and deep learning.    

    

  主讲人简介:李文雯,美国亚利桑那州立大学(ASU)地理科学与城市规划学院副教授(终身教授)、博士生导师、 ASU可持续发展中心资深科学家、计算机系联合教授、ASU巴雷特荣誉学院荣誉教授,网络信息基础设施与智能计算实验室主任(http://cici.lab.asu.edu),空间信息科学研究中心副主任。主要研究方向为WebGIS、时空大数据、 人工智能与知识挖掘,以及科学可视化。她的研究成果广泛的应用在多个地球科学和社会科学领域,包括土地利用和土地覆盖变化,北极气候变化,生物量估算,地貌特征提取等等。在ASU任职期间,作为第一项目负责人主持多项美国自然科学基金项目,个人总获奖金额超过170万美元。并在诸多知名国际GIS及遥感杂志上发表论文70余篇,其中SCI论文40篇。同时,她也是美国自然科学基金(NSF)颁发的杰出青年教授奖得主(得奖年份:2015),该奖项为NSF为美国Tenure-Track教授颁发的最高荣誉。 

    

  主持人:秦承志 研究员 

  时  间:2018年5月30号(周三)上午9:30 

  地  点:中科院地理资源所2321会议室 

    

  欢迎广大科研人员和研究生光临交流! 


附件下载: