A crowdsourcing tool has been developed for Φ-lab to support activities where manual interpretation of EO data is required. Based on Pybossa, an Open Source framework for crowdsourcing, the tool allows users to freely publish EO-related crowdsourcing projects, or contribute to already published activities.
An example activity that has been released on this platform is crowdsourcing for archaeological prospection. This forms part of a Φ-lab initiative that attempts to prototype an automatic methodology for retrieving archaeological cropmarks through a combination of human interpretation and machine learning. Users are presented with tiles of high-resolution optical EO data, and are asked to identify vegetation patterns as proxies of buried structures.
Results show that a redundancy of three independent interpretations of each tile is sufficient to yield accurate classifications, with many detected buried archaeological features including roads, buildings and even urban areas. Once a critical mass of training data is available, the intention is to use it to train a machine learning model to scale automatic detections over a wider area, which currently only includes the region surrounding the city of Rome.
The pressure of development is putting the cultural heritage of many areas at risk. Results of this activity will demonstrate efficient alternatives to the conventional ground-based techniques of rescue archaeology
