Is it technically and economically feasible to use commercial remote sensing imagery for crowdsourcing to validate geo-data?

A case study was conducted within the ESA Project "Thematic Exploitation Platform URBAN (TEP Urban)," in cooperation with Pallas Ludens and cloudeo.

Earth Observation imagery is the basis for nearly every geoinformation service. Such large datasets, often in the range of terabytes, require quality assurance, which is generally done based on sample data; similarly, training data are needed to train automated classification algorithms. For geoinformation services like the European Copernicus program, quality assurance through validation is a prerequisite. The collection of sample data is based on reference image data or ground truth data; validation is typically performed by geo-experts in a very elaborate, time-consuming, and costly process, especially for large-area and semantically complex products.

One possible approach to this problem is to use non-experts and crowdsourcing. To explore this possibility, the European Space Agency launched a case study within their project "Thematic Exploitation Platform URBAN" (TEP Urban); the German Aerospace Center DLR (Deutsches Zentrum für Luft- und Raumfahrt) subcontracted the companies Pallas Ludens and cloudeo to answer the question: Is it technically and economically feasible to use commercial remote sensing imagery for crowdsourcing to validate geo-data?

For this study, the DLR Global Urban Footprint (GUF) product was used as an example for a classification layer. Pallas Ludens provided their expertise in crowdsourcing techniques; cloudeo was chosen for their valuable experience with making EO data accessible to many users. In addition to brokering access to data from their content partners, cloudeo also evaluated possible business models for the particular use case of crowdsourcing.

Pallas Ludens was tasked with setting up the technical framework for a state-of-the-art crowd-based collection of sample data, consisting of multi-resolution geo-imagery; cloudeo provided their Service API to set up the interface for the application in one homogenous web service access. About the validation of buildings or group of buildings within GUF, real color imagery was pre-selected, with a pixel size equal to or better than 15 m:
• Very high-resolution imagery ( resolution better than 1m)
• High-resolution imagery (resolution better than 2.5m)
• Visually improved imagery with 15m resolution.
The cloudeo Service API creates homogeneous access to the data and simplifies access for the crowdsourcing application. Besides, the embedded metering supports pay per use and application-specific business models.

To control for this case study's cost, cloudeo selected services from SI Imaging and Airbus Defence and Space, and PlanetObserver, who offered data for free for the study. California and New Delhi were chosen as areas of interest; a detailed analysis of crowdsourcing performance was executed on the California test site. The New Delhi demo area was used as the second test site.

cloudeo holds agreements with some providers, like PlanetObserver, which allows archive data with temporal licenses for one day, one, or several months for reduced costs. This model is very well suited for crowdsourcing. Due to a smart selection of areas and tasking by the application, a fast access time to the image will be, in most cases, sufficient to get the results. Thus, the application may even benefit from a one-day license. This daily license was offered for PlanetObersver's PlanetSAT 15 imagery. cloudeo hosted the global data set of PlanetSAT 15 and provided this data set as an international web service. The pricing depends on the number of users and the area. It is independent of the number of map views. The pricing seems suitable for the low-cost crowdsourcing application: For a minimum fee of 500 Euro, nearly 30,000 sq. km. can be accessed for one month.