All of you have heard a few things about machine learning, which has become very popular these days. Machine learning is used in many areas such as autonomous devices, face-speech-object recognition, anomaly detection, market analysis, forecasting, etc. Thanks to the development of machine learning algorithms and technological developments, it is used and tested especially in many subjects of visual fields these days. Also, today many social media applications such as facebook, twitter, instagram, pinterest etc. use back-plane machine learning algorithms.
Machine learning is also used in the interpretation of satellite images except of all these areas. Satellite images have different resolution characteristics and multi-layer structure (HSI). Therefore, different methods have been tried and developed in order to take advantage of these properties during training and in the extraction of features. In this article, I would like to share with you the slide presentation that I have prepared for a graduate lesson at this subject and I have prepared an article in a format similar to IEEE format as a very brief summary of the different researches and their proposal methods.
Machine learning is a product of the work of many disciplines such as probability and stochastic processes, linear algebra, computer science, biology etc. Therefore, if you are encountering with this for the first time, these presentations and articles will not be enough for you to get a clear idea. If I can spare time, I can write a more comprehensive article on this subject. You can access a lot of qualified tutorials and articles about this topic on the internet. Also, you can access my suggestions and other references, that i used in slide-show and at the paper from the end of this article. Before looking at these, I strongly recommend that you watch this video, which is told by David Wiesel of the experiments done by Wiesel & Hubel in order to understand our brain’s working logic and our visual perception ability and this famous experiment of them on cats visual cortex.
Machine learning is computationaly expensive processes. Generally, these processes done by python language by the data scientists and machine learning developers, because of the speed and having various libraries. There are tons of library for the machine learning like theano, caffe, tensorflow etc. You can use one of them for your ML problems. Also you can code in matlab or java for your ML applications.
It is a good one for beginner (both mathematical intuation and coding): https://www.coursera.org/learn/machine-learning
This is more intermediate level : http://cs231n.github.io/
A video series of mathematical intuation behind ML : Youtube Link
This is ML practical applications with Python (coding) : Youtube Link
The References :
- Deep Convolutional Neural Networks for Hyperspectral Image Classification Wei Hu, Yangyu Huang, Li Wei, Fan Zhang, and Hengchao Li, Hindawi Publishing Corporation Journal of Sensors Volume 2015, Article ID 258619, 12 pages
- Urban Land Use and Land Cover Classification Using Remotely Sensed SAR Data through Deep Belief Networks Qi Lv, Yong Dou, Xin Niu, Jiaqing Xu, Jinbo Xu, and Fei Xia, Hindawi Publishing Corporation Journal of Sensors Volume 2015, Article ID 538063, 10 pages
- Stationary Aircraft Detection From Satellite Images Polat and C. Yildiz/ IU-JEEE Vol. 12(2), (2012), 1523-1528
- John A. Richards Remote Sensing Digital Image Analysis Book, Springer-Verlag 3rd printing 1995