Remote Sensing Environmental Changes

 

 

  NSF-INRIA Collaborative  

 

 

   

Objective

The increasing number of satellites, dedicated to the understanding of the Earth system and the effects of natural and human-induced changes on the global environment, gather a large number of data for scientists and decision makers. The purpose of the associated scientific missions is to provide the governments and the public with the basis of well founded environmental and resources management policy formulation.

The observations of different processes reflecting natural and human-induced changes generate a large number of data that have to be processed for identifying or detecting patterns of changes in the Earth system. Naturally, a large number of sciences are involved in understanding the Earth system, and one common framework is the use of the observation data gathered by the satellites as inputs for environmental modeling. A large number of satellite acquisitions are in fact temporal image sequences acquired by the sensors. On another hand, Computer Vision techniques are widely used for detecting and characterizing changes in image sequences. However, these methods are mainly developed for video sequences and rely on a set of assumptions such as changes in gray level values are due to relative motion of the observer v.s. the objects in the scene. This is not often the case for environmental remote sensing, where change in gray level sometimes reflects change in object properties (Land Surface Temperature, vegetation cycle...).

Temporal changes can be first understood as a local variation. For example, texture-based monitoring of forest and crop changes reflects, at a seasonal scale, the vegetation evolution and at an annual scale, the variation of the forest type and/or quality. These changes in texture allow to characterize the growth stage of the vegetation health (tree’s dryness, vegetation stress...). Temporal changes in remote sensing can also depicts motion and deformation of structures. For example, detecting vortices and characterizing front evolution in Sea Surface Temperature or atmospheric images (clouds) allows to identify patterns of ocean/atmospheric circulation.

The purpose of this proposal is defining a network of collaboration between teams with different specificities at USC and INRIA to address the problem of change detection and characterization in environmental image sequences. We will focus on two different environmental problems where the detection and characterization of temporal changes is crucial:

  • vegetation and forestry monitoring, and
  • oceanography

The teams involved in this effort have developed methods and new approaches for addressing the underlying problems in computer vision and have strong links with environmental monitoring and applications. We believe that defining the present network of collaborations will help us in leveraging on each team achievements in order to evaluate the existing potentials among us for deriving a system allowing to automatically detect and characterize changes in environmental image sequences.

People