Remote Sensing & Machine Learning

Course Aims

• To build competence in using current remote sensing techniques and sensors in earth surface applications;
• To explain the physics of optical sensing methods and the consequences of system and process characteristics to information extraction capabilities and reliability;
• To practice the processing of remote sensing data in a structured and repeatable manner (Python scripting) in both a desktop- and cloud-based image processing environment.
• To gain knowledge and understanding of machine learning methods and their applications on geographic data.

During the course you will develop and train the following skills:

• Giving academic oral presentations about an applied remote sensing topic.
• Written reporting about image processing and interpretation.
• Analyze and interpret various types of satellite images using the theoretical knowledge acquired during the lectures.
• Hands on use of advanced image processing software to process, interpret, classify and analyze a range of different earth observation images.

The student is expected to:

• Understand the fundamentals of image acquisition and analysis and its applications;
• Be able to analyze and interpret remote sensing information in their spatial and temporal contexts;
• Be able to apply image processing and machine learning methods and effectively use build-in or online documentation to compose their own analyses.
• To critically evaluate information resources and remote sensing products for scientific applications.

Remote sensing, or Earth observation, is a fast developing technical toolkit of exceptional importance for all geo-disciplines. Earth observation is widely used to study the dynamics of system Earth and deliver important input in global change models, vegetation condition and changes, and at regional level for modeling catchment and erosion processes. Remote sensing enables the collection of information about the spatial distribution of objects at the Earth surface such as vegetation, soil or rock types, snow, surface water, and to identify object properties (primary productivity, soil mineral composition, terrain conditions) and to investigate their temporal changes (seasonal or long-term). A wide range of sensors (optical, thermal, radar, lidar) are now orbiting the Earth or are available in aircrafts or unmanned platforms. In this course we build explore advanced techniques for information extraction from imagery by hands-on exercises.

Machine learning methods (sometimes also addressed as deep learning or AI) include advanced statistical algorithms that can learn from data and generalize to unseen data. These methods have many applications in remote sensing image processing, and data analysis in general. This course will introduce the foundational background of the machine learning techniques commonly applied in earth surface and water research. We will also work hands on with specifically those methods suited for spatial data such as random forests, convolutional neural networks, and other deep learning models which surpass many previous applied classification or regression approaches in performance.