Developing and Evaluating Simplified Tools for Image Processing in a Problem-Based Learning Environment for Earth Observation

Abstract

Abstract Earth observation is an interdisciplinary field of application. This makes it ideal for supporting natural science education in schools. Understanding satellite data can enable students to see more than just “beautiful images”. They can gain insights into an exciting field of application of many natural- and computer science fundamentals. In recent years, interactive lessons have been produced by our working group. They address the prospects and challenges connected with the teaching of earth observation topics in schools. The principles upon which the lessons are built promote a problem-based style of learning combined with a low impact in extraneous cognitive load for the students. Advances in web technology have made it necessary for interactive lessons to be redesigned and rethought. Contemporary web technologies have been selected and combined into an application framework to fulfill the requirements of our design principles and put a greater emphasis on usability during the content creation process. To allow offline usage of the learning modules it was tested whether this can be achieved using strictly client-side image processing. This article shows that, based on the technology stack used, lessons can be created that enable pupils to analyze remote sensing data in a much-simplified way. Using those simplified methods, a classification with an overall accuracy of 78.66% was achieved. The comparison of another simplified classification using just threshold values resulted in up to 89% probability to predict the outcome of a professionally produced dataset, making the simplified methods applicable in real-world examples presented in problem-based teaching scenarios.

Cite this work

No public copy is stored for this record, so only the metadata is cited.

Stored only when both papers are in the corpus — 5 citations exist upstream.

Papers that cite the same references — the relational fallback while embeddings are still backfilling.