The French National Research Agency Projects for science

Voir cette page en français

ANR funded project

JCJC - SIMI 2 - Science informatique et applications (JCJC SIMI 2)
Edition 2013


ASTERIX


Spatio-Temporal Analysis by Recognition within Complex Images for Remote Sensing of Environment

Spatio-Temporal Analysis by Recognition within Complex Images for Remote Sensing of Environment
The goal of the project and its originality is to bring new methods, algorithms, software in the field of image analysis and machine learning in order to support recognition within complex images, by explicitly dealing with the specificity of remote sensing complex images. In this context, the main challenges are related to high dimensionality, heterogeneity, volume and spatio-temporal behaviour of images.

General objective of the project and main issues raised
Following the growth of multisource data with high spatial, spectral, and temporal resolutions, the problem of complex image information mining in remote sensing of environment becomes a great challenge, with many potential applications raising. However, there is no or only a few methodological frameworks for dealing with data with multiple spatial and temporal scales: recognition methods are most often straight applications of standard classification and modelling methods. Besides, dealing with spatial and temporal neighbourhood, with various kinds of data, is expected to improve significantly resulting performances.
The goal of the project and its originality is to bring new methods, algorithms, software in the field of image analysis and machine learning in order to support recognition within complex images, by explicitly dealing with the specificity of remote sensing complex images. In this context, the main challenges are related to high dimensionality, heterogeneity, volume and spatio-temporal behaviour of images.
Besides methodological achievements supporting the development of the state-of-the-art in image processing and machine learning in the context of recognition within complex images, expected results from the ASTERIX project consist in a set of concrete solutions to crucial problems in remote sensing of environment, and specifically for coastal and mountains environments. More precisely, considered applications are related to the dynamic of environmental objects, which help to understand coastal evolution, and the dynamic of ash tree colonization in an agricultural mountain landscape.

Methods and technologies used
The ASTERIX project aims to deliver advanced techniques for spatio-temporal analysis from complex data, in the context of image analysis and understanding in remote sensing of environment. To do so, several methodological problems in image analysis and machine learning have to be tackled to face the complexity of processed data. More precisely, such a complexity intrinsic to remote sensing will be addressed through three research directions:
- complexity brought by heterogeneous and multi-source images ;
- complexity brought by massive and high-dimensional data ;
- complexity brought by additional the temporal information.
These three directions lead to a project organization into three methodology-oriented tasks.
Furthermore, the project will also address application issues in remote sensing of environment. Two specific landscapes will be considered: coastal areas and mountains. Both are considered in a unique task that will benefit from the various methodological contributions.

Results

Mid-term results mainly consist in methodological developments, among which:
- some new methods for feature selection (WHISPERS 2014 et JSTARS 2015), morphological description of hyperspectral images (ICIP 2014), supervised description of hyperspectral images with tree-based representations (IIM 2014 et WHISPERS 2014), description of elevation data (DSM) with tree-based representations (IGARSS 2015), and kernel-based classification of hierarchical image representations (GbR 2015) ;
- a morphological approach for road detection (PRRS 2014) ;
- fusion of optical and radar satellite image time series (Master thesis) ;
- analysis and classification of hyperspectral images with manifold learning (JSTARS 2014), non-negative matrix factorization (Machine Learning 2014) ;
- active learning of time series (Master thesis) ;
- comparison of classification methods for time series (Master thesis), new approach using bag of words (ECML/PKDD 2015) ;
- domain adaptation based on optimal transport (ECML/PKDD 2014) and anomaly detection based on dimensionality reduction (ECML/PKDD 2014).

Outlook

The ASTERIX project addresses issues that are of high interest for a wide community, both from the methodological side (in image analysis and machine learning) and the applications in remote sensing of environment. ASTERIX results are expected to be compared with the national and international state-of-the-art. At the national level, this will be achieved through national scientific days. Special sessions and workshop in major conferences will allow gaining interest from a wide audience. Such events could also lead to special issues to be edited in international journals (e.g. in remote sensing).
The project will be promoted through a dedicated website. Besides a description of the issues addressed in the project, the website will host some prototypes demonstrating the potential impact of the solutions to be elaborated in the project. Such solutions could also be disseminated as open-source software components. OSUR will be involved in order to ease and accelerate the design and production of software tools, through the image platform it provides.

Scientific outputs and patents

International journal papers:
- N. Courty, X. Gong, J. Vandel, T. Burger. SAGA: Sparse And Geometry-Aware non-negative matrix factorization through non-linear local embedding. Machine Learning, 97(1-2):205-226, 2014
- L. Chapel, T. Burger, N. Courty, S. Lefèvre. PerTurbo Manifold Learning Algorithm for Weakly Labeled Hyperspectral Image Classification. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, 7(4):1070-1078, 2014
- M. Fauvel, C. Dechesne, A. Zullo, F. Ferraty. Fast Forward Feature Selection of Hyperspectral Images for Classification With Gaussian Mixture Models. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, 2015

International conference papers:
- F. Merciol, L. Chapel, S. Lefèvre. Hyperspectral Image Representation through alpha-Trees. IIM 2014
- M. Fauvel, A. Zullo, F. Ferraty. Nonlinear Parsimonious Feature Selection for the Classification of Hyperspectral Images. WHISPERS 2014
- S. Lefèvre, L. Chapel, F. Merciol. Hyperspectral Image Classification from Multiscale Description with Constrained Connectivity and Metric Learning. WHISPERS 2014
- L. Courtrai, S. Lefèvre. Road Network Extraction from Remote Sensing using Region-based Mathematical Morphology. PRRS 2014
- L. Chapel, C. Friguet, Anomaly detection with score functions based on the reconstruction error of the kernel PCA. ECML/PKDD 2014
- N. Courty, R. Flamary, D. Tuia. Domain adaptation with regularized optimal transport. ECML/PKDD 2014
- E. Aptoula, N. Courty S. Lefèvre. An end-member based ordering relation for the morphological description of hyperspectral images. ICIP 2014
- Y. Cui, L. Chapel, S. Lefèvre. A subpath kernel for learning hierarchical image representations. GbR 2015
- F. Merciol, S. Lefèvre. Fast building extraction by multiscale analysis of digital surface models. IGARSS 2015
- A. Bailly, S. Malinowski, R. Tavenard, T. Guyet, L. Chapel. Bag-of-Temporal-SIFT-Words for Time Series Classification. ECML/PKDD 2015

Partners

IRISA Institut de Recherche en Informatique et Systèmes Aléatoires

ANR grant: 275 979 euros
Beginning and duration: octobre 2013 - 48 mois

Submission abstract

Following the growth of multisource data with high spatial, spectral, and temporal resolutions, the problem of complex image information mining in remote sensing of environment becomes a great challenge, with many potential applications raising. However, there is no or only a few methodological frameworks for dealing with data with multiple spatial and temporal scales: recognition methods are most often straight applications of standard classification and modelisation methods. Besides, dealing with spatial and temporal neighborhood, with various kinds of data, is expected to improve significantly resulting performances.

The goal of the ASTERIX project (Spatio-Temporal Analysis by Recognition within Complex Images for Remote Sensing of Environment) and its originality is to bring new methods, algorithms, softwares in the field of image analysis and machine learning in order to support recognition within complex image, by explicitly dealing with the specificity of remote sensing complex images. In this context, main challenges are related to high dimensionality, heterogeneity, volume and spatio-temporal behaviour of images.

Besides methodological achievements supporting the development of the state-of-the-art in image processing and machine learning in the context of recognition within complex images, expected results from the ASTERIX project consist in a set of concrete solutions to crucial problems in remote sensing of environment, and especially in two environment: coastal and montains. More precisely, applications considered are related to the dynamic of environmental objects which help to understand coastal evolution, the dynamic of ash tree colonization in an agricultural mountain landscape, and the dynamic of geological process.

 

ANR Programme: JCJC - SIMI 2 - Science informatique et applications (JCJC SIMI 2) 2013

Project ID: ANR-13-JS02-0005

Project coordinator:
Monsieur Sébastien Lefèvre (Institut de Recherche en Informatique et Systèmes Aléatoires)

Project web site: http://anr-asterix.irisa.fr

 

Back to the previous page

 

The project coordinator is the author of this abstract and is therefore responsible for the content of the summary. The ANR disclaims all responsibility in connection with its content.