机器学习与GIS研究进展

              机器学习与GIS研究进展

王少华

    本文简述了机器学习在GIS领域的新进展。

    目前机器学习研究很热，机器学习（ML），包括人工神经网络（ANN）和支持向量机（SVM）等，为智能地理环境数据分析、处理和可视化提供了极其重要的工具。机器学习是对地理统计学等传统技术的重要补充。机器学习在空间数据处理中扮演重要角色，本文介绍了针对地理空间数据的几种现代机器学习的应用，包括环境数据的区域分类，连续环境和污染数据的制图，基于自动算法，优化（设计/从新设计）监测网络等。详细可参考文献1。

     空间变换神经网络成果的发表（文献3），绝对值得用惊艳来描述。这篇文章是Google旗下的新锐AI公司DeepMind的四位剑桥Phd研究员发表的成果。卷积神经网络（CNN）已经被证实可以训练一个能力强大的分类模型，但与传统的模式识别方法相似，它也会受到数据在空间上多样性的影响。这篇Paper提出了一种叫作空间变换神经网络（Spatial Transform Networks， STN）,该网络不须要关键点的标定，可以根据分类或者其它任务自适应地将数据进行空间变换和对齐（包括平移、缩放、旋转以及其它几何变换等）。在输入数据在空间差别较大的状况下，这个网络能够加在现有的卷积网络中，提升分类的准确性。该论文中的案例包括在手写文字识别、街景数字识别、鸟类分类以及共定位等方面。不久以后，能够预料这篇文章的引用将飙升！参考文献4就是基于此方法技术的学位论文。

      

        

参考文献：

1.Kanevski, M., A. Pozdnukhov, and V. Timonin. "Machine learning algorithms for geospatial data. Applications and software tools." (2008).

2.Ostermann, F. "Hybrid geo-information processing: Crowdsourced supervision of geo-spatial machine learning tasks." Proceedings of the 18th AGILE International Conference on Geographic Information Science, Lisbon, Portugal. 2015.

3.Jaderberg, Max, Karen Simonyan, and Andrew Zisserman. "Spatial transformer networks." Advances in Neural Information Processing Systems. 2015.

4.Chianucci, Daniel. "Detection in Aerial Images Using Spatial Transformer Networks." (2016).

5.Xie, Michael, et al. "Transfer learning from deep features for remote sensing and poverty mapping." arXiv preprint arXiv:1510.00098 (2015).

6.Jean, Neal, et al. "Combining satellite imagery and machine learning to predict poverty." Science 353.6301 (2016): 790-794.

7.Allen, Chris, et al. "Applying GIS and Machine Learning Methods to Twitter Data for Multiscale Surveillance of Influenza." PloS one 11.7 (2016): e0157734.

8.Weyand, Tobias, Ilya Kostrikov, and James Philbin. "Planet-photo geolocation with convolutional neural networks." European Conference on Computer Vision. Springer International Publishing, 2016.

9.Costea, Dragos, and Marius Leordeanu. "Aerial image geolocalization from recognition and matching of roads and intersections." arXiv preprint arXiv:1605.08323 (2016).

10.Sharma, Prafull, Michel Schoemaker, and David Pan. "Automated Image Timestamp Inference Using Convolutional Neural Networks." (2016).

11.Brahmbhatt, Samarth, and James Hays. "DeepNav: Learning to Navigate Large Cities." arXiv preprint arXiv:1701.09135 (2017).

12.Gupta, Saurabh, et al. "Cognitive Mapping and Planning for Visual Navigation." arXiv preprint arXiv:1702.03920 (2017).

13.Ma, Xiaolei, et al. "Large-scale transportation network congestion evolution prediction using deep learning theory." PloS one 10.3 (2015): e0119044.

14.Zhao, Wenzhi, et al. "On combining multiscale deep learning features for the classification of hyperspectral remote sensing imagery." International Journal of Remote Sensing 36.13 (2015): 3368-3379.

 

 

 

转载请注明连接：https://my.oschina.net/u/2338162/blog/843723