The Experts below are selected from a list of 17916 Experts worldwide ranked by ideXlab platform

Morin David - One of the best experts on this subject based on the ideXlab platform.

  • Estimation et suivi de la ressource en bois en France métropolitaine par valorisation des séries multi-temporelles à haute résolution spatiale d'images optiques (Sentinel-2) et radar (Sentinel-1, ALOS-PALSAR)
    HAL CCSD, 2020
    Co-Authors: Morin David
    Abstract:

    The estimation and monitoring of forest resources and carbon stocks are major issues for wood industry and public bodies. Forests play an important role in national and international plans for climate change mitigation (carbon storage, climate regulation, biodiversity, renewable energy). In temperate forests, monitoring is done at two different levels: on one hand, at local level, in small areas by the acquisition of many measures of forest structure parameters, and, on the other hand, by statistics at national level or in large administrative areas that are provided annually by public bodies. Temperate forests are highly anthropogenic (high spatial variability and fragmentation of stands), so there is currently a strong need for a more refined and Regular maps of forest resources in these regions. Optical and radar satellite images provide information on the state of vegetation, tree structure and spatial organization of forests. In an exceptional context of free global availability, diversity, and quality of images with high spatial and temporal resolution, the aim of this PhD work is to set up the methodological bases for a generic and semi-automatic production of forest parameters Mapping (biomass, diameter, height, etc.). We have assessed the potential of Sentinel-1 (C-band radar), Sentinel-2 (optical) time series, and ALOS2-PALSAR2 (radar, L-band) annual mosaics to estimate forest structure parameters. These satellite data are combined, using supervised learning algorithms and field measurements, to construct models for estimating aboveground biomass (AGB), mean tree diameter (DBH), height, basal area and tree density. These models can then be spatially applied over the entire territory by using satellite images, providing thus continuous information on the spatial resolution of the images used (10 to 20 meters). This approach has been conceived and tested on four study sites with different forest species and structural and environmental properties: the inner and the dune zone of the Landes forest (maritime pines), the Orléans forest (oak and Scots pines), and the forest of Saint-Gobain (oaks, hornbeams and beeches). The investigated issues are the satellite data to be used, the selection of explanatory variables, the choice of regression algorithms and their parameterization, the differentiation of forest types and the spatialization of forest parameter estimates. The primitives derived from satellite data provide information on the optical properties of soil and vegetation, the spatial organization of trees, the structure and volume of live wood of crowns and trunks. The use of nonlinear multivariate regression algorithms allows to obtain forest parameter estimates with relative error performance in the order of 15 to 35 % for the basal area (~ 2.8 to 5.9 m2/ha) depending on forest types, 5 to 20 % for height (~ 1.3 to 3 m), and 5 to 25 % for DBH (~ 1.5 to 8 cm). The results highlight the improvement by combining several types of satellite data (optical, multi-frequency radar and spatial texture indexes), as well as the importance of differentiating forest types for the construction of models. This high-resolution, Regular Mapping of the forest resource is very promising to help improving the monitoring and policy of territorial and national strategies for the timber sector, biodiversity and carbon storage.L'estimation et le suivi du carbone et de la ressource forestière sont des enjeux majeurs pour la gestion des territoires. Les forêts ont un rôle important dans les plans nationaux et internationaux pour l'atténuation du changement climatique (stockage du carbone, régulation du climat, biodiversité, énergies renouvelables). Dans les forêts tempérées, de nombreuses mesures des paramètres de structure forestière sont acquises sur des petites zones, des statistiques au niveau national ou sur de larges zones administratives sont délivrées annuellement par les organismes gouvernementaux. Les forêts tempérées sont fortement anthropisées (forte variabilité spatiale et fractionnement des peuplements), il y a actuellement un besoin fort d'une spatialisation plus fine et continue des ressources forestières dans ces régions. Les images satellitaires optiques et radar apportent des informations sur l'état de la végétation, la structure des arbres et l'organisation spatiale des forêts. Dans un contexte exceptionnel de disponibilité mondiale et gratuite, de diversité, de qualité des images à haute résolution spatiale et temporelle, le travail de thèse a pour objectif de mettre en place les bases méthodologiques et scientifiques pour une production nationale semi-automatique d'une cartographie des paramètres forestiers (biomasse, diamètre, hauteur, etc.). Nous avons évalué le potentiel des séries temporelles Sentinel-1 (radar en bande C), Sentinel-2 (optique), et des mosaïques annuelles ALOS2-PALSAR2 (radar, bande L) pour estimer les paramètres de structure forestière. Ces données satellitaires ont été combinées à l'aide d'algorithmes d'apprentissage supervisé et de mesures terrain pour construire des modèles d'estimation de la biomasse, du diamètre moyen des arbres (DBH), de la hauteur et d'autres paramètres de structure. Ces modèles peuvent ensuite être spatialisés sur l'ensemble du territoire à l'aide des images satellitaires, et fournir une information continue à la résolution spatiale des images utilisées (10 à 20 mètres). Notre approche a été conçue et testée sur quatre sites d'étude avec des essences forestières et des propriétés structurales et environnementales différentes : la zone intérieure et la zone dunaire de la forêt des Landes (pins maritimes), la forêt d'Orléans (chênes et pins sylvestres), et la forêt de Saint-Gobain (chênes, charmes et hêtres). Les principaux développements portent sur les données satellitaires à utiliser, la sélection des variables explicatives, le choix des algorithmes de régression et leur paramétrisation, la différenciation des types de forêt et la cartographie des estimations de paramètres forestiers. Les primitives issues des données satellitaires fournissent des informations sur les propriétés optiques du sol et de la végétation, l'organisation spatiale des arbres, la structure et le volume de bois vivant des houppiers et des troncs. L'utilisation d'algorithmes de régression multivariée non-linéaire permet d'obtenir des estimations des paramètres forestiers avec des performances en termes d'erreur relative de l'ordre de 15 à 35 % pour la surface terrière (~2.8 à 5.9 m2/ha) selon les types de forêt, 5 à 20 % pour la hauteur (~1.3 à 3 m), et de 5 à 25 % pour le DBH (~1.5 à 8 cm). Les résultats montrent l'apport de la combinaison de plusieurs types de données satellitaires (optique, radar multi-fréquence et indices de texture spatiale) ainsi que l'importance de différencier les types de forêt pour la construction des modèles. L'application des modèles sur les images satellitaires permet de produire des cartes à haute résolution spatiale de ces paramètres forestiers, utilisables de l'échelle locale à l'échelle régionale/nationale

  • Estimation et suivi de la ressource en bois en France métropolitaine par valorisation des séries multi-temporelles à haute résolution spatiale d'images optiques (Sentinel-2) et radar (Sentinel-1, ALOS-PALSAR)
    2020
    Co-Authors: Morin David
    Abstract:

    L'estimation et le suivi du carbone et de la ressource forestière sont des enjeux majeurs pour la gestion des territoires. Les forêts ont un rôle important dans les plans nationaux et internationaux pour l'atténuation du changement climatique (stockage du carbone, régulation du climat, biodiversité, énergies renouvelables). Dans les forêts tempérées, de nombreuses mesures des paramètres de structure forestière sont acquises sur des petites zones, des statistiques au niveau national ou sur de larges zones administratives sont délivrées annuellement par les organismes gouvernementaux. Les forêts tempérées sont fortement anthropisées (forte variabilité spatiale et fractionnement des peuplements), il y a actuellement un besoin fort d'une spatialisation plus fine et continue des ressources forestières dans ces régions. Les images satellitaires optiques et radar apportent des informations sur l'état de la végétation, la structure des arbres et l'organisation spatiale des forêts. Dans un contexte exceptionnel de disponibilité mondiale et gratuite, de diversité, de qualité des images à haute résolution spatiale et temporelle, le travail de thèse a pour objectif de mettre en place les bases méthodologiques et scientifiques pour une production nationale semi-automatique d'une cartographie des paramètres forestiers (biomasse, diamètre, hauteur, etc.). Nous avons évalué le potentiel des séries temporelles Sentinel-1 (radar en bande C), Sentinel-2 (optique), et des mosaïques annuelles ALOS2-PALSAR2 (radar, bande L) pour estimer les paramètres de structure forestière. Ces données satellitaires ont été combinées à l'aide d'algorithmes d'apprentissage supervisé et de mesures terrain pour construire des modèles d'estimation de la biomasse, du diamètre moyen des arbres (DBH), de la hauteur et d'autres paramètres de structure. Ces modèles peuvent ensuite être spatialisés sur l'ensemble du territoire à l'aide des images satellitaires, et fournir une information continue à la résolution spatiale des images utilisées (10 à 20 mètres). Notre approche a été conçue et testée sur quatre sites d'étude avec des essences forestières et des propriétés structurales et environnementales différentes : la zone intérieure et la zone dunaire de la forêt des Landes (pins maritimes), la forêt d'Orléans (chênes et pins sylvestres), et la forêt de Saint-Gobain (chênes, charmes et hêtres). Les principaux développements portent sur les données satellitaires à utiliser, la sélection des variables explicatives, le choix des algorithmes de régression et leur paramétrisation, la différenciation des types de forêt et la cartographie des estimations de paramètres forestiers. Les primitives issues des données satellitaires fournissent des informations sur les propriétés optiques du sol et de la végétation, l'organisation spatiale des arbres, la structure et le volume de bois vivant des houppiers et des troncs. L'utilisation d'algorithmes de régression multivariée non-linéaire permet d'obtenir des estimations des paramètres forestiers avec des performances en termes d'erreur relative de l'ordre de 15 à 35 % pour la surface terrière (~2.8 à 5.9 m2/ha) selon les types de forêt, 5 à 20 % pour la hauteur (~1.3 à 3 m), et de 5 à 25 % pour le DBH (~1.5 à 8 cm). Les résultats montrent l'apport de la combinaison de plusieurs types de données satellitaires (optique, radar multi-fréquence et indices de texture spatiale) ainsi que l'importance de différencier les types de forêt pour la construction des modèles. L'application des modèles sur les images satellitaires permet de produire des cartes à haute résolution spatiale de ces paramètres forestiers, utilisables de l'échelle locale à l'échelle régionale/nationale. La possibilité de produire une cartographie régulière à haute résolution de la ressource forestière est très prometteuse pour aider à améliorer la mise en place et le suivi des stratégies territoriales et nationales pour la filière bois, la biodiversité et le stockage du carbone.The estimation and monitoring of forest resources and carbon stocks are major issues for wood industry and public bodies. Forests play an important role in national and international plans for climate change mitigation (carbon storage, climate regulation, biodiversity, renewable energy). In temperate forests, monitoring is done at two different levels: on one hand, at local level, in small areas by the acquisition of many measures of forest structure parameters, and, on the other hand, by statistics at national level or in large administrative areas that are provided annually by public bodies. Temperate forests are highly anthropogenic (high spatial variability and fragmentation of stands), so there is currently a strong need for a more refined and Regular maps of forest resources in these regions. Optical and radar satellite images provide information on the state of vegetation, tree structure and spatial organization of forests. In an exceptional context of free global availability, diversity, and quality of images with high spatial and temporal resolution, the aim of this PhD work is to set up the methodological bases for a generic and semi-automatic production of forest parameters Mapping (biomass, diameter, height, etc.). We have assessed the potential of Sentinel-1 (C-band radar), Sentinel-2 (optical) time series, and ALOS2-PALSAR2 (radar, L-band) annual mosaics to estimate forest structure parameters. These satellite data are combined, using supervised learning algorithms and field measurements, to construct models for estimating aboveground biomass (AGB), mean tree diameter (DBH), height, basal area and tree density. These models can then be spatially applied over the entire territory by using satellite images, providing thus continuous information on the spatial resolution of the images used (10 to 20 meters). This approach has been conceived and tested on four study sites with different forest species and structural and environmental properties: the inner and the dune zone of the Landes forest (maritime pines), the Orléans forest (oak and Scots pines), and the forest of Saint-Gobain (oaks, hornbeams and beeches). The investigated issues are the satellite data to be used, the selection of explanatory variables, the choice of regression algorithms and their parameterization, the differentiation of forest types and the spatialization of forest parameter estimates. The primitives derived from satellite data provide information on the optical properties of soil and vegetation, the spatial organization of trees, the structure and volume of live wood of crowns and trunks. The use of nonlinear multivariate regression algorithms allows to obtain forest parameter estimates with relative error performance in the order of 15 to 35 % for the basal area (~ 2.8 to 5.9 m2/ha) depending on forest types, 5 to 20 % for height (~ 1.3 to 3 m), and 5 to 25 % for DBH (~ 1.5 to 8 cm). The results highlight the improvement by combining several types of satellite data (optical, multi-frequency radar and spatial texture indexes), as well as the importance of differentiating forest types for the construction of models. This high-resolution, Regular Mapping of the forest resource is very promising to help improving the monitoring and policy of territorial and national strategies for the timber sector, biodiversity and carbon storage

  • Estimation and monitoring of wood resources in France by using high resolution time series of optical and radar images
    2020
    Co-Authors: Morin David
    Abstract:

    L'estimation et le suivi du carbone et de la ressource forestière sont des enjeux majeurs pour la gestion des territoires. Les forêts ont un rôle important dans les plans nationaux et internationaux pour l'atténuation du changement climatique (stockage du carbone, régulation du climat, biodiversité, énergies renouvelables). Dans les forêts tempérées, de nombreuses mesures des paramètres de structure forestière sont acquises sur des petites zones, des statistiques au niveau national ou sur de larges zones administratives sont délivrées annuellement par les organismes gouvernementaux. Les forêts tempérées sont fortement anthropisées (forte variabilité spatiale et fractionnement des peuplements), il y a actuellement un besoin fort d'une spatialisation plus fine et continue des ressources forestières dans ces régions. Les images satellitaires optiques et radar apportent des informations sur l'état de la végétation, la structure des arbres et l'organisation spatiale des forêts. Dans un contexte exceptionnel de disponibilité mondiale et gratuite, de diversité, de qualité des images à haute résolution spatiale et temporelle, le travail de thèse a pour objectif de mettre en place les bases méthodologiques et scientifiques pour une production nationale semi-automatique d'une cartographie des paramètres forestiers (biomasse, diamètre, hauteur, etc.). Nous avons évalué le potentiel des séries temporelles Sentinel-1 (radar en bande C), Sentinel-2 (optique), et des mosaïques annuelles ALOS2-PALSAR2 (radar, bande L) pour estimer les paramètres de structure forestière. Ces données satellitaires ont été combinées à l'aide d'algorithmes d'apprentissage supervisé et de mesures terrain pour construire des modèles d'estimation de la biomasse, du diamètre moyen des arbres (DBH), de la hauteur et d'autres paramètres de structure. Ces modèles peuvent ensuite être spatialisés sur l'ensemble du territoire à l'aide des images satellitaires, et fournir une information continue à la résolution spatiale des images utilisées (10 à 20 mètres). Notre approche a été conçue et testée sur quatre sites d'étude avec des essences forestières et des propriétés structurales et environnementales différentes : la zone intérieure et la zone dunaire de la forêt des Landes (pins maritimes), la forêt d'Orléans (chênes et pins sylvestres), et la forêt de Saint-Gobain (chênes, charmes et hêtres). Les principaux développements portent sur les données satellitaires à utiliser, la sélection des variables explicatives, le choix des algorithmes de régression et leur paramétrisation, la différenciation des types de forêt et la cartographie des estimations de paramètres forestiers. Les primitives issues des données satellitaires fournissent des informations sur les propriétés optiques du sol et de la végétation, l'organisation spatiale des arbres, la structure et le volume de bois vivant des houppiers et des troncs. L'utilisation d'algorithmes de régression multivariée non-linéaire permet d'obtenir des estimations des paramètres forestiers avec des performances en termes d'erreur relative de l'ordre de 15 à 35 % pour la surface terrière (~2.8 à 5.9 m2/ha) selon les types de forêt, 5 à 20 % pour la hauteur (~1.3 à 3 m), et de 5 à 25 % pour le DBH (~1.5 à 8 cm). Les résultats montrent l'apport de la combinaison de plusieurs types de données satellitaires (optique, radar multi-fréquence et indices de texture spatiale) ainsi que l'importance de différencier les types de forêt pour la construction des modèles. L'application des modèles sur les images satellitaires permet de produire des cartes à haute résolution spatiale de ces paramètres forestiers, utilisables de l'échelle locale à l'échelle régionale/nationale.The estimation and monitoring of forest resources and carbon stocks are major issues for wood industry and public bodies. Forests play an important role in national and international plans for climate change mitigation (carbon storage, climate regulation, biodiversity, renewable energy). In temperate forests, monitoring is done at two different levels: on one hand, at local level, in small areas by the acquisition of many measures of forest structure parameters, and, on the other hand, by statistics at national level or in large administrative areas that are provided annually by public bodies. Temperate forests are highly anthropogenic (high spatial variability and fragmentation of stands), so there is currently a strong need for a more refined and Regular maps of forest resources in these regions. Optical and radar satellite images provide information on the state of vegetation, tree structure and spatial organization of forests. In an exceptional context of free global availability, diversity, and quality of images with high spatial and temporal resolution, the aim of this PhD work is to set up the methodological bases for a generic and semi-automatic production of forest parameters Mapping (biomass, diameter, height, etc.). We have assessed the potential of Sentinel-1 (C-band radar), Sentinel-2 (optical) time series, and ALOS2-PALSAR2 (radar, L-band) annual mosaics to estimate forest structure parameters. These satellite data are combined, using supervised learning algorithms and field measurements, to construct models for estimating aboveground biomass (AGB), mean tree diameter (DBH), height, basal area and tree density. These models can then be spatially applied over the entire territory by using satellite images, providing thus continuous information on the spatial resolution of the images used (10 to 20 meters). This approach has been conceived and tested on four study sites with different forest species and structural and environmental properties: the inner and the dune zone of the Landes forest (maritime pines), the Orléans forest (oak and Scots pines), and the forest of Saint-Gobain (oaks, hornbeams and beeches). The investigated issues are the satellite data to be used, the selection of explanatory variables, the choice of regression algorithms and their parameterization, the differentiation of forest types and the spatialization of forest parameter estimates. The primitives derived from satellite data provide information on the optical properties of soil and vegetation, the spatial organization of trees, the structure and volume of live wood of crowns and trunks. The use of nonlinear multivariate regression algorithms allows to obtain forest parameter estimates with relative error performance in the order of 15 to 35 % for the basal area (~ 2.8 to 5.9 m2/ha) depending on forest types, 5 to 20 % for height (~ 1.3 to 3 m), and 5 to 25 % for DBH (~ 1.5 to 8 cm). The results highlight the improvement by combining several types of satellite data (optical, multi-frequency radar and spatial texture indexes), as well as the importance of differentiating forest types for the construction of models. This high-resolution, Regular Mapping of the forest resource is very promising to help improving the monitoring and policy of territorial and national strategies for the timber sector, biodiversity and carbon storage

Jurgen Teich - One of the best experts on this subject based on the ideXlab platform.

  • Regular Mapping for coarse grained reconfigurable architectures
    International Conference on Acoustics Speech and Signal Processing, 2004
    Co-Authors: Frank Hannig, Hritam Dutta, Jurgen Teich
    Abstract:

    Similar to programmable devices such as processors or micro controllers, reconfigurable logic devices can also be built as software, by programming the configuration of the device. In this paper, we present an overview of constraints which have to be considered when Mapping applications to coarse-grained reconfigurable architectures. The application areas of most of these architectures address computational-intensive algorithms like video and audio processing or wireless communication. Therefore, reconfigurable arrays are in direct competition with DSP processors which are traditionally used for digital signal processing. Hence, existing Mapping methodologies are closely related to approaches from the DSP world. They try to employ pipelining and temporal partitioning but they do not exploit the full parallelism of a given algorithm and the computational potential of typically 2D arrays. We present a first case study for Mapping Regular algorithms onto reconfigurable arrays by using our design methodology which is characterized by loop parallelization in the polytope model. The case study shows that our Regular Mapping methodology may lead to highly efficient implementations taking the constraints of the architecture into account.

  • Mapping of Regular nested loop programs to coarse grained reconfigurable arrays constraints and methodology
    International Parallel and Distributed Processing Symposium, 2004
    Co-Authors: Frank Hannig, Hritam Dutta, Jurgen Teich
    Abstract:

    Summary form only given. Apart from academic, recently more and more commercial coarse-grained reconfigurable arrays have been developed. Computational intensive applications from the area of video and wireless communication seek to exploit the computational power of such massively parallel SoCs. Conventionally, DSP processors are used in the digital signal processing domain. Thus, the existing compilation techniques are closely related to approaches from the DSP world. These approaches employ several loop transformations, like pipelining or temporal partitioning, but they are not able to exploit the full parallelism of a given algorithm and the computational potential of a typical 2-dimensional array. In this paper, (i) we present an overview of constraints which have to be considered when Mapping applications to coarse-grained reconfigurable arrays, (ii) we present our design methodology for Mapping Regular algorithms onto massively parallel arrays which is characterized by loop parallelization in the polytope model, and (Hi), in a first case study, we adapt our design methodology for targeting reconfigurable arrays. The case study shows that the presented Regular Mapping methodology may lead to highly efficient implementations taking into account the constraints of the architecture.

Frank Hannig - One of the best experts on this subject based on the ideXlab platform.

  • Regular Mapping for coarse grained reconfigurable architectures
    International Conference on Acoustics Speech and Signal Processing, 2004
    Co-Authors: Frank Hannig, Hritam Dutta, Jurgen Teich
    Abstract:

    Similar to programmable devices such as processors or micro controllers, reconfigurable logic devices can also be built as software, by programming the configuration of the device. In this paper, we present an overview of constraints which have to be considered when Mapping applications to coarse-grained reconfigurable architectures. The application areas of most of these architectures address computational-intensive algorithms like video and audio processing or wireless communication. Therefore, reconfigurable arrays are in direct competition with DSP processors which are traditionally used for digital signal processing. Hence, existing Mapping methodologies are closely related to approaches from the DSP world. They try to employ pipelining and temporal partitioning but they do not exploit the full parallelism of a given algorithm and the computational potential of typically 2D arrays. We present a first case study for Mapping Regular algorithms onto reconfigurable arrays by using our design methodology which is characterized by loop parallelization in the polytope model. The case study shows that our Regular Mapping methodology may lead to highly efficient implementations taking the constraints of the architecture into account.

  • Mapping of Regular nested loop programs to coarse grained reconfigurable arrays constraints and methodology
    International Parallel and Distributed Processing Symposium, 2004
    Co-Authors: Frank Hannig, Hritam Dutta, Jurgen Teich
    Abstract:

    Summary form only given. Apart from academic, recently more and more commercial coarse-grained reconfigurable arrays have been developed. Computational intensive applications from the area of video and wireless communication seek to exploit the computational power of such massively parallel SoCs. Conventionally, DSP processors are used in the digital signal processing domain. Thus, the existing compilation techniques are closely related to approaches from the DSP world. These approaches employ several loop transformations, like pipelining or temporal partitioning, but they are not able to exploit the full parallelism of a given algorithm and the computational potential of a typical 2-dimensional array. In this paper, (i) we present an overview of constraints which have to be considered when Mapping applications to coarse-grained reconfigurable arrays, (ii) we present our design methodology for Mapping Regular algorithms onto massively parallel arrays which is characterized by loop parallelization in the polytope model, and (Hi), in a first case study, we adapt our design methodology for targeting reconfigurable arrays. The case study shows that the presented Regular Mapping methodology may lead to highly efficient implementations taking into account the constraints of the architecture.

Hritam Dutta - One of the best experts on this subject based on the ideXlab platform.

  • Regular Mapping for coarse grained reconfigurable architectures
    International Conference on Acoustics Speech and Signal Processing, 2004
    Co-Authors: Frank Hannig, Hritam Dutta, Jurgen Teich
    Abstract:

    Similar to programmable devices such as processors or micro controllers, reconfigurable logic devices can also be built as software, by programming the configuration of the device. In this paper, we present an overview of constraints which have to be considered when Mapping applications to coarse-grained reconfigurable architectures. The application areas of most of these architectures address computational-intensive algorithms like video and audio processing or wireless communication. Therefore, reconfigurable arrays are in direct competition with DSP processors which are traditionally used for digital signal processing. Hence, existing Mapping methodologies are closely related to approaches from the DSP world. They try to employ pipelining and temporal partitioning but they do not exploit the full parallelism of a given algorithm and the computational potential of typically 2D arrays. We present a first case study for Mapping Regular algorithms onto reconfigurable arrays by using our design methodology which is characterized by loop parallelization in the polytope model. The case study shows that our Regular Mapping methodology may lead to highly efficient implementations taking the constraints of the architecture into account.

  • Mapping of Regular nested loop programs to coarse grained reconfigurable arrays constraints and methodology
    International Parallel and Distributed Processing Symposium, 2004
    Co-Authors: Frank Hannig, Hritam Dutta, Jurgen Teich
    Abstract:

    Summary form only given. Apart from academic, recently more and more commercial coarse-grained reconfigurable arrays have been developed. Computational intensive applications from the area of video and wireless communication seek to exploit the computational power of such massively parallel SoCs. Conventionally, DSP processors are used in the digital signal processing domain. Thus, the existing compilation techniques are closely related to approaches from the DSP world. These approaches employ several loop transformations, like pipelining or temporal partitioning, but they are not able to exploit the full parallelism of a given algorithm and the computational potential of a typical 2-dimensional array. In this paper, (i) we present an overview of constraints which have to be considered when Mapping applications to coarse-grained reconfigurable arrays, (ii) we present our design methodology for Mapping Regular algorithms onto massively parallel arrays which is characterized by loop parallelization in the polytope model, and (Hi), in a first case study, we adapt our design methodology for targeting reconfigurable arrays. The case study shows that the presented Regular Mapping methodology may lead to highly efficient implementations taking into account the constraints of the architecture.

Winter P. - One of the best experts on this subject based on the ideXlab platform.

  • Design and performance of an in-vacuum, magnetic field Mapping system for the Muon g-2 experiment
    'IOP Publishing', 2021
    Co-Authors: Corrodi S., De Lurgio P., Flay D., Grange J., Hong R., Kawall D., Oberling M., Ramachandran S., Winter P.
    Abstract:

    The E989 Muon $g-2$ experiment at Fermilab aims to measure the anomalous magnetic moment, $a^{}_\mu$, of the muon with a precision of 140 parts-per-billion. This requires a precise measurement of both the anomalous spin precession frequency, $\omega^{}_a$, and the average magnetic field in terms of the shielded proton Larmor frequency, $\omega'^{}_p$. The measurement of $\omega'^{}_p$ with a total systematic uncertainty of 70 parts-per-billion involves a combination of various NMR probes. There are 378 probes in fixed locations constantly monitoring field drifts. A water-based probe provides the calibration. A crucial element for the multi-step measurement of $\omega'^{}_p$ is the Regular Mapping of the magnetic field over the muon storage region. The former E821 experiment at Brookhaven employed an in-vacuum field Mapping system equipped with 17 NMR probes, which was developed by the University of Heidelberg. We have refurbished and upgraded this system with new probes and electronics. The upgrades include a new communication scheme incorporating time-division multiplexing to separate the important NMR reference clock from the data communication. The addition digitization of the NMR signals replaced the hardware-implemented zero-crossing counting of the E821 system. The digitized signals offer new capabilities in the NMR frequency analysis and its related systematic uncertainties. While the mechanical systems that move the field mapper around the ring have been mostly refurbished, the motion control system was completely replaced with a custom-built electronics centered around a commercial Galil motion controller. Both the field Mapping NMR system and its motion control were successfully commissioned at Fermilab and have been in reliable operation during the first data taking periods. This article provides details of the upgrades of the field mapper and its performance

  • Design and performance of an in-vacuum, magnetic field Mapping system for the Muon g-2 experiment
    2020
    Co-Authors: Corrodi S., De Lurgio P., Flay D., Grange J., Hong R., Kawall D., Oberling M., Ramachandran S., Winter P.
    Abstract:

    The E989 Muon g-2 experiment at Fermilab aims to measure the anomalous magnetic moment, a$_\mu$, of the muon with a precision of 140 parts-per-billion. This requires a precise measurement of both the anomalous spin precession frequency, $\omega_a$, and the average magnetic field in terms of the equivalent, free proton Larmor frequency, $\omega_p$. The measurement of $\omega_p$ with a total systematic uncertainty of 70 parts-per-billion involves a combination of various NMR probes. There are 378 probes in fixed locations constantly monitoring field drifts. A water-based probe provides the calibration in terms of $\omega_p$. A crucial element for the multi-step measurement of $\omega_p$ is the Regular Mapping of the magnetic field over the muon storage region. The former E821 experiment at Brookhaven employed an in-vacuum field Mapping system equipped with 17 NMR probes, which was developed by the University of Heidelberg. We have refurbished and upgraded this system with new probes and electronics. The upgrades include a new communication scheme incorporating time-division multiplexing to separate the important NMR reference clock from the data communication. The addition digitization of the NMR signals replaced the hardware-implemented zero-crossing counting of the E821 system. The digitized signals offer new capabilities in the NMR frequency analysis and its related systematic uncertainties. While the mechanical systems that move the field mapper around the ring have been mostly refurbished, the motion control system was completely replaced with a custom-built electronics centered around a commercial Galil motion controller. Both the field Mapping NMR system and its motion control were successfully commissioned at Fermilab and have been in reliable operation during the first data taking periods. This article provides details of the upgrades of the field mapper and its performance.Comment: To be submitted to JINS