PEPS - High Added-Value Image Processing as a Service
Bricier, Aurelien1; Dedieu, Gerard2; Hebrard, Dominique3; Guzzonato, Eric1
1CS-SI, FRANCE; 2CESBIO, FRANCE; 3CETE SO, FRANCE

The "PEPS" project aims at bridging the gap between abundant spatial resources and their access and use by public and/or private users. It will provide to the greatest number remote sensing and spatial imaging assets through simple tools based on proven methods. It will be a major technology transfer carrier from public research laboratories with knowledge that matches the users specific needs. Moreover; since 2010, the evaluation of urban planning policies has been a legal obligation and has to be done every three years. The tools and data from space technologies in the field of earth observation matches particularly with this institutional need. Thanks to the specifications of the new and coming satellites, wider areas can be observed, with higher space resolution and better acquisition frequency. The interest in creating such tool with a simple interface for institutional users comes with great expectation.

This project is funded by CNES in the field of the last call of innovative applications in 2011. It involves three partners: CS-SI, CETE and CESBIO and potential end-users. The underlying target is to prototype a flexible and scalable client-server tool set with centralized data management and computing power. In order to maximize the adoption rate and simplify the roolout, a lightweight dedicated web client will guarantee the user a popularized access. In order to fulfill its ambitions, the project relies on already successful, well-tried open source software components and a set of robust and validated methodological processes. In fine, the user will gain access to high value-added remote sensing products through easy-to-use processing chains without the need of an image processing background.
PEPS originality and interest is not bound to the final service it will provide but also the methodological approach which aims at capitalizing both past experiences in remote sensing and the interaction between tool designers and end users. Within this context, a first step would be to focus on the appropriation of remote sensing practices by the departments of the French Ministery of Ecology. This step should also be driven by the final goal to deploy a scalable service accordingly to the needs of potential public policies or private users in a wide thematic spectrum. Another major ambition of this project is the popularization of specialized and high-end remote sensing processing chains through ergonomic complexity-free user interface, erasing the necessity of image processing background for the user.

In this project, three major axes will be developed. The first one is to integrate various open source software components into a robust, yet interoperable service platform. The interoperability will be guaranteed by an OGC standard compatibility: WFS and WMS for data exchange and visualization and WPS to trigger remote image processing chains. This OGC standard compliance will also provide the platform scalability to various use cases and data sources. This software basis will host the results of the second development line: the implementation of high-end image analysis processes applied to remote sensing images and based on already existing well-tried algorithms. The specifications of the processing tools as well as the user interface are built in a collaborative way with users of the public services in charge of Environment. Taking advantages from crowdsourced user supervision or heterogeneous data sources and types, the resulting service will benefit from the user community implication and resources. The final line of development, and probably the most challenging, is the popularization of high-end image processing algorithms and complex concepts through minimal algorithm parameterization, simplified and well documented use cases and intuitive and ergonomic user interfaces. All these restrictions should be considered accordingly to the users specific needs. It implies complete comprehension of both the user needs and expectations in terms of understanding simplicity. After a prospective survey, the needs formulated by future users are to extract specific landscape objects included within the followings types:

Atifical Areas

  • Urban Areas
  • Sprawlng Urban Fabric
  • Dense Urban Fabric
  • Agricultural Areas
    Woodlands
  • Woods
  • Deciduous Woodlands
  • Conifer Woodlands
  • Hedge Network
    Waters
  • Inlands Waters
  • Waterways
  • Lakes

    In the short term, PEPS will implement a reliable classification tool able to run large scale classification to produce land cover mapping according to a subset or all of these classes. It will also implement a basic GIS for geographic product creation and edition. Specific mask creation, dedicated to waters or woodlands for example, will also be possible. Thus, automatic land cover map production tools will be offered to end users as a service through means that do not require remote sensing background, and can benefit from crowdsourced experience.