Grégoire Maillet's research while affiliated with Institut national de l’information géographique et forestière and other places

La futaie de pin maritime du Massif des Landes de Gascogne fait l’objet d’une sylviculture dynamique ; les dommages occasionnés par les tempêtes Martin (1999) et Klaus (2009) puis par des ravageurs (scolytes) ont entrainé une forte diminution de la ressource en bois. Cela a accentué la demande de la filière forêt-bois aquitaine pour un suivi permanent, précis et spatialisé de celle-ci. L’étude Pinastéréo ouvre des perspectives nouvelles en la matière grâce au modèle numérique de surface (MNS) obtenu par corrélation d’images stéréoscopiques Pléiades. Sa combinaison avec le modèle numérique de terrain BD Alti® fournit le modèle de hauteur de canopée (MNHC). La précision des MNS et MNHC a été évaluée avec des mesures de référence (LiDAR aérien, placettes de levers dendrométriques) ; elle s'est avérée satisfaisante. Il est montré que la hauteur dominante des peuplements de pin maritime, leur volume aérien total et leur biomasse peuvent être modélisés à partir du MNHC. Une carte de la hauteur dominante a alors été produite sur un site de référence ; celle de la biomasse reste à faire. Les faibles incertitudes observées sur les estimations de hauteur et de volume sont très prometteurs pour une généralisation de la méthode au Massif landais ou même à d'autres futaies régulières résineuses.

The maritime pine stands of 'Landes de Gascogne' forest is the object of a dynamic silviculture; Martin (1999) and Klaus (2009) windstorms have provoked a lot of damages and a great decrease of the resource. A permanent monitoring of the resource and its mapping has been reinforced. In this context, the study Pinastéréo explore some perspectives by testing a digital photogrammetry technique. Digital surface model and canopy height model have been realized by images Pleiades matching and by using the digital terrain model BD Alti®. Thanks to reference measures (aerial LiDAR, field plot measurements), the accuracy of DSM and CHM have been estimated. The top height of the coniferous stands, their aboveground volume and biomass were calibrated from extracted variables of CHM and then top height has been mapped for a specific zone. The satisfying accuracy of the models of the top height and of the aboveground volume of the coniferous stands is promising. A generalization all over the 'Landes de Gascogne' forest and a use for others coniferous even-aged stands is now allowed.

Durant ces dernières décennies, l'imagerie par télédétection aéroportée des milieux urbains reste un enjeu scientifique majeur. L'arrivée d'une nouvelle génération de caméras aéroportées permettant d'effectuer des acquisitions à très haute résolution spatiale et possédant un meilleur rapport signal sur bruit ouvre de nouvelles perspectives pour extraire l'information de ces environnements. En effet, la principale limitation de la plupart des méthodes de traitement des images dans ces milieux est due à la grande variété des matériaux qui les compose et à la géométrie dense et complexe qui les caractérise et peut induire de larges zones d'ombre. La campagne (UMBRA) a pour but d'acquérir un nouveau jeu de données de référence représentatif de deux villes françaises distinctes en vue de développer de nouvelles méthodes de correction atmosphérique au soleil et à l'ombre, de concevoir de futures caméras superspectrales dédiées à des applications de télédétection urbaine, et d'étudier des applications pour ``ré-éclairer'' des scènes à partir d'images urbaines terrestres. Ce papier détaille cette campagne ainsi que les prétraitements qui ont été effectués sur les différentes données acquises. De premiers résultats sont également présentés.

Remote sensing imagery of urban environments from airborne acquisitions still remains a major scientific issue over the last decades. As urban areas are characterized by a high variety of materials and many shadowed areas, most of the processing methods are limited. Moreover, due to the high spatial resolution and better signal-to-noise ratio of the new generation of airborne sensors, this opens the way to new developments to improve the extraction of information from such environments. This campaign (UMBRA) aims at acquiring a new reference dataset representative of two different French cities to develop a new atmospheric correction method in sunlit and shaded urban areas, to design a future generation of superspectral cameras system dedicated to geomatic applications, and to study relighting applications from terrestrial urban image data. This paper details this experiment and the preprocessing done on the different acquired data. Some first results are presented.optical properties ; experimental studies ; geomatics ; atmospheric correction ; extraction ; signal-to-noise ratio ; spatial resolution ; urban areas ; urban environment ; imagery ; remote sensing ; trees ; Sun ; materials ; ecosystems ;

During the past decades 3D building extraction has been a constant topic of research for several universities and cartographic or photogrammetric institutes because there was a great demand for 3D city models in commercial applications. These applications move from electromagnetic propagation for telecommunication to more demanding simulations for acoustic, urban planning, virtual and augmented reality. In most cases, 3D complete roof shapes are required. Year after year radiometric quality and geometric accuracy of new high resolution satellites keep on improving so much so satellite imagery become a real potential solution for the production of such 3D models. This present paper studies the potential of the CNES (French Spatial Agency) Pleiades-HR system, ready to launch in 2007, for the 3D city models production in the case of 2D ground maps availability. First because this Pleiades-HR system, able to acquire tristereoscopic high-resolution panchromatic images, seems well adapted for 3D applications. Secondly because 2D ground maps are now available in many countries, especially in dense urban areas, and are very useful for the automation of the reconstruction process. For the present evaluation very fine image simulations provided by the CNES are used, in addition to 2D ground plans that come from the French cadastral database. The 3D models have been produced with an operational framework able to reconstruct buildings from optical images and 2D ground plans. A 3D reference has been acquired from aerial images with a resolution of 25cm. This reference allows to evaluate from qualitative and quantitative points of view the 3D reconstruction. Finally, the quality of 3D city models produced from the Pleiades-HR images will be discussed in terms of levels of details, geometric accuracy, exhaustiveness and productivity.

The Pleiades system, of which the first platform of the constellation shall be ready for launch in early 2009, is the new generation of high resolution optical sensor satellites developed by CNES, the French space agency. With its ability to provide photogrammetrists with instantaneous colour image stereo-pairs of 20km swath and 70cm resolution acquired with B/H ratios ranging from 0.15 to 2, the very agile and precisely pointing Pleiades platforms should open in many countries a new era for the updating of high scale maps, by probably avoiding heavy aerial survey campaigns in many cases. In the framework of a cooperation agreement between CNES and IGN-France (IGN-F), designed to prepare an operational use of these images, several experimentations and simulations are being carried out, in order to verify the ability of the system to fulfill the technical and operational requirements of the map-making processes, in comparison with those presently based on digital aerial photography. Among others, present studies particularly focus on : • the geometric and radiometric quality assessment of the planned system products (sensor level, orthoimages and mosaics). • the capacity assessment of the Pleiades constellation to meet the IGN-F cartographic mission needs in terms of image volume for the coverage of the national territory, whether in monoscopic or stereoscopic mode. • the potential of stereo-and three-fold stereoscopic image simulations for the extraction of buildings in 3D, after adapting algorithms primarily dedicated to higher resolution aerial images. After an overview of the main characteristics of the Pleiades system, the paper aims at presenting the first results of these studies, which will be of great importance, together with the economic side, in the decision process of IGN-F to transfer, or not, part of the supplying of fresh images from the aerial source to the satellite source.

The paper focuses on comparing accuracies obtained with photogrammetry and laser scanning in building extraction and partly com-pares results obtained with various methods. The objective of the EuroSDR Building Extraction comparison was to evaluate the quality, accuracy, feasibility and economical aspects of semi-automatic building extraction based on photogrammetric techniques with the emphasis on commercial and/or operative systems, semi-automatic and automatic building extraction techniques based on high density laser scanner data and semi-automatic and automatic building extraction techniques based on integration of laser scan-ner data and aerial images (hybrid technique). The project consists of three test sites by Finnish Geodetic Institute (FGI), namely Senaatti, Hermanni and Espoonlahti and one test site by Institut Geographique National (IGN), namely Amiens. For each test site fol-lowing data was provided to the partners: aerial images, camera calibration and image orientation information, ground control point coordinates and jpg images of point locations (not for Amiens), laser scanner data and cadastral map vectors of selected buildings. Participants were requested to create the vectors of 3D city models. 3D-models were obtained from 11 participants. Paper confirms with experiments that laser scanning is more suitable in deriving building heights, extracting planar roof faces and ridges of the roof whereas the photogrammetry and aerial images are more suitable in building outline and length determination. CyberCity, Stuttgart and TerraScan (performed by ICC) solutions provided the highest accuracy. There seemed to be a higher variation in the quality of other models depending on test site or remotely sensed information.

The paper focuses on comparing accuracies obtained with photogrammetry and laser scanning in building extraction. The objective of the EuroSDR Building Extraction comparison was to evaluate the quality, accuracy, feasibility and economical aspects of semi-automatic building extraction based on photogrammetric techniques with the emphasis on commercial and/or operative systems, semi-automatic and automatic building extraction techniques based on high density laser scanner data and semi-automatic and automatic building extraction techniques based on integration of laser scanner data and aerial images. The project consists of three test sites by the Finnish Geodetic Institute (FGI), namely Senaatti, Hermanni and Espoonlahti, and one test site by the Institut Geographique National (IGN), namely Amiens. For each test site following data was provided to the partners: aerial images, camera calibration and image orientation information, ground control point coordinates and jpg images of point locations (not for Amiens), laser scanner data and cadastral map vectors of selected buildings. Participants were requested to create vectors of 3D city models. 3D-models were obtained from 11 participants. The paper confirms with experiments that laser scanning is superior in deriving building heights, extracting planar roof faces and ridges of the roof, whereas photogrammetry and aerial images are superior in building outline and length determination.

3D city models production is a difficult and expensive problem. Automatic processes in the production lines seem to be necessary to produce 3D city models at reasonable costs. This paper presents a 3D reconstruction software platform, based on several algorithms developed in the MATIS laboratory, which deals with aerial or space images and cadastral maps. Due to landscape complexities and disparities, the choice of a unique and efficient production way is not possible. In our approach, the operator manages his strategy and combines as he wants the different algorithm modules. Step by step a standard scenario seems to be identified. Tests have been carried out with two different input data contexts: first, using IGN digital camera images, and then, with Pleiades-HR simulations.

Even though many efficient building reconstruction methods have been published so far, automatic processes are still missing for the production of 3D city models. This gap between research results and effective production tools seems mainly due to a lack of cooperation between all these methods. Building reconstruction in dense urban areas is such a complex problem that it is hopeless to look for an universal solution that could efficiently reconstruct every building, from a simple gabled roof to a whole cathedral. In this paper we first present a general strategy to combine several automatic and user-assisted processes to produce three-dimensional city models in a real production context using aerial images and 2D ground maps. The strategy aims at taking advantage of each algorithm to be sure they complement each other. Then we illustrate this strategy by describing our operational implementation and its results. In this implementation, the first step is a combination of two automatic algorithms: a model driven and a data driven process. In a second step, user-assisted tools supplement the reconstruction: some are adapted to precise landscape features while others are more generic.