Geo-reCAPTCHA: Crowdsourcing large amounts of geographic information from earth observation data

- Adapting reCAPTCHA idea to create geographic information from earth observation data.
- Geo-reCAPTCHA can be solved for building digitizations on average in 19.2 s.
- Geo-reCAPTCHA provides overall average accuracy of building digitizations of 82.2%.
- Geo-reCAPTCHA has the potential to be a reasonable alternative to reCAPTCHA.
- Proof-of-Concept for crowd sourcing of geoinformation via web-based micro-mapping tasks.

The reCAPTCHA concept provides a large amount of valuable information for various applications. First, it provides security, e.g., for a form on a website, by means of a test that only a human could solve. Second, the effort of the user for this test is used to generate additional information, e.g., digitization of books or identification of house numbers. In this work, we present a concept for adapting the reCAPTCHA idea to create user-generated geographic information from earth observation data, and the requirements during the conception and implementation are depicted in detail. Furthermore, the essential parts of a Geo-reCAPTCHA system are described, and afterwards transferred, to a prototype implementation. An empirical user study is conducted to investigate the Geo-reCAPTCHA approach, assessing time and quality of the resulting geographic information. Our results show that a Geo-reCAPTCHA can be solved by the users of our study on building digitization in a short amount of time (19.2 s on average) with an overall average accuracy of the digitizations of 82.2%. In conclusion, Geo-reCAPTCHA has the potential to be a reasonable alternative to the typical reCAPTCHA, and to become a new data-rich channel of crowdsourced geographic information.