The Sentinel-2 (S2) satellite constellation offers new opportunities for monitoring the Earth. However, the raw data is difficult to use due to cloud cover and other limitations (10m resolution limited for analysis at fine scale, 16-bits representation that prevents from large scale visualization for example). As a consequence, several platforms have emerged to synthesize the data and offer easy-to-use products over the past years.Among them, Nimbo is a free, cloud-based platform that provides users with access to monthly images of the entire Earth. The platform uses deep neural networks to remove clouds from satellite images, colorizing images and providing users with a clear view of the Earth’s surface.Providing homogeneous images makes it easier to develop large-scale tasks, in particular by creating reliable and robust datasets for various applications. This is illustrated in this paper, through two important applications: change detection and super-resolution.

Generally caused by extreme events, landslides cause severe landscape modifications and may endanger local population. It is important to be able to map them in order to better understand landscape evolution. 3D LiDAR point clouds (PCs) are a relevant choice compared to 2D imagery to directly sense ground shape modification under vegetated areas. Most of the studies propose to rely on rasterization of PCs, or multistep semi-automatic process with tedious manual results refinement in the 3D PCs. In this study, we aim at experimenting a deep learning method to directly extract landslide sources and deposits from raw 3D PCs. To this end, we train an Encoder Fusion SiamKPConv network, designed for 3D PCs change detection, for the specific task of landslides identification in PCs acquired before and after Kaikōura earthquake (New-Zealand). The experimental results (93.87% of accuracy) show the relevance of this model.

In this paper, we propose a deep neural network architecture to estimate physical and optical properties of water bodies from bathymetric lidar waveforms. Although essential to the understanding of coastal and inland waters dynamics, this task remains challenging due to the complexity of waveform processing. Here, we use convolutional encoders to estimate seven parameters without the need for pre-processing, iterations, or existing measurements of the target properties. Using a data simulator based on radiative transfer models, the network is trained to be robust to a wide range of physical and acquisition settings. On simulated data, the results show the ability of the method to retrieve relevant Kd, depth, and bottom position estimates even for low signal-to-noise ratios in which the water bottom component is particularly weak and the water column difficult to identify.

Remote sensing techniques are increasingly employed for early detection of ground deformation and landslide warning systems. The combination of vast remote sensing data and advancements in machine learning algorithms has led to significant progress in landslide detection and mapping.This study proposes an innovative neural network architecture for landslide detection. The network utilizes a fusion of optical images (RGB) and Digital Elevation Models (DEMs) to enhance accuracy. Additionally, attention layers and Mixup techniques are incorporated to further improve the model’s performance. Given the limited availability of training data, the proposed network was trained on a publicly accessible dataset, specifically the established Bijie landslide dataset located in China. We demonstrate that using an efficient training strategy on this dataset allows us to pretrain a network that can be easily fine-tuned for a different site exhibiting landslides at various scales. In practice, we introduce a newly developed dataset covering the southwestern part of Hokkaido, Japan, which experienced a landslide event in 2018. This dataset is freely available on the internet. The results demonstrate the effectiveness of the proposed architecture compared to existing methods, highlighting the benefits of our pretrained model.

In this study, we tackle the challenge of identifying plant species from ultra high resolution (UHR) remote sensing images. Our approach involves introducing an RGB remote sensing dataset, characterized by millimeter-level spatial resolution, meticulously curated through several field expeditions across a mountainous region in France covering various landscapes. The task of plant species identification is framed as a semantic segmentation problem for its practical and efficient implementation across vast geographical areas. However, when dealing with segmentation masks, we confront instances where distinguishing boundaries between plant species and their background is challenging. We tackle this issue by introducing a fuzzy loss within the segmentation model. Instead of utilizing one-hot encoded ground truth (GT), our model incorporates Gaussian filter refined GT, introducing stochasticity during training. First experimental results obtained on both our UHR dataset and a public dataset are presented, showing the relevance of the proposed methodology, as well as the need for future improvement.

Change detection is an important task that rapidly identifies modified areas, particularly when multi-temporal data are concerned. In landscapes with a complex geometry (e.g., urban environment), vertical information is a very useful source of knowledge that highlights changes and classifies them into different categories. In this study, we focus on change segmentation using raw three-dimensional (3D) point clouds (PCs) directly to avoid any information loss due to the rasterization processes. While deep learning has recently proven its effectiveness for this particular task by encoding the information through Siamese networks, we investigate herein the idea of also using change information in the early steps of deep networks. To do this, we first propose to provide a Siamese KPConv state-of-the-art (SoTA) network with hand-crafted features, especially a change-related one, which improves the mean of the Intersection over Union (IoU) over the classes of change by 4.70%. Considering that a major improvement is obtained due to the change-related feature, we then propose three new architectures to address 3D PC change segmentation: OneConvFusion, Triplet KPConv, and Encoder Fusion SiamKPConv. All these networks consider the change information in the early steps and outperform the SoTA methods. In particular, Encoder Fusion SiamKPConv overtakes the SoTA approaches by more than 5% of the mean of the IoU over the classes of change, emphasizing the value of having the network focus on change information for the change detection task. The code is available at https://github.com/IdeGelis/torch-points3d-SiamKPConvVariants.

This paper is interested with consistent rendering of 16-bits SENTINEL-2 images into 8-bits ones in order to produce visually sound global products (useful for basemap for example). Though the colorization of a single 16-bits data can efficiency be done with histogram stretching techniques for example, applying a unique transformation able to generate consistent data with enough details/contrasts whatever the content of the original image remains challenging. We propose here to train a neural network on a wide variety of images that have been manually enhanced to derive a unique model able to consistently recolor images and generate sound global products without mosaic effects.

This paper is concerned with the change detection issue in remote sensing images. This problem is not trivial since the notion of change depends on the application. Moreover, classical supervised deep learning methods have to deal with the limited amount of labelled data available. Based on existing deep learning techniques that exploit unsupervised clustering to assign labels to entire images, we adapt them to the change detection problem by using siamese backbones and extracting pixel-wise results. As fully unsupervised experiments lead to unstable results, we suggest "low supervision" strategy composed of a warm-up stage with few labeled data able to drive the following unsupervised learning through reliable solutions. Preliminary experiments show reliable change maps.

This study aims at recovering above-ground biomass information from ultra-high resolution UAV RGB-NIR orthophotos. We focus on a realistic scenario where a limited number of training samples for a landscape with heterogeneous herbaceous vegetation is given. Consequently, we explore different machine learning methods explicitly addressing the limitations of small training samples and compare their predictions quantitatively and qualitatively. Our results show that random forest models perform similarly well to deep learning models. While simpler machine learning models may, therefore, still be preferable, our study also points the way to promising architectures and regularisation techniques for deep learning approaches. Beyond vegetation cover, accurate regression of other variables, including vegetation height, volume and biomass remains a difficult task regardless of the model choice.

This paper is concerned with the decomposition of bathymetric lidar waveforms. Because of the presence of water, processing such data remains a challenge since water impacts their shape and signal-to-noise ratio, depending in particular on the associated turbidity. In this paper, we explore the use of attentive autoencoders to decompose bathymetric waveforms and recover their air/water interface, water column, and water bottom components simultaneously, without relying on assumptions about the impulse or target surface nature. On simulated waveforms, the method achieves lower decomposition error than existing approaches, handling overlapping echoes of very shallow waters and weak returns in deeper water. This opens to attractive strategies to process real bathymetric waveforms.

Recently, deep convolutional networks have made great progress on the task of super resolution, i.e. reconstructing images with finer spatial resolution. However, although the reconstructions are visually impressive, they may lack physical consistency. This aspect is sought in remote sensing, where the resolution of satellite imagery (e.g. Sentinel-2) may be too coarse to characterize the physical structure and dynamics of certain landscapes. Through the study of flooding dynamics in wet grasslands, we propose a super resolution approach that allows deriving fine resolution patterns that are visually realistic and physically exploitable. This approach is based on an architecture, Fusion-UNet, allowing the fusion of multispectral data with a digital terrain model (DTM) associated with a loss function combining content, structure and segmentation losses. Our results show that this model can precisely predict water levels while restituting the fine structure of the landscape. This approach allows to refine the production of hydrological and ecological indicators to define the state of the ecosystem.

According to the United Nations, 70% of earth population is going to live in cities by 2050. Given this fast urban evolution, urban monitoring is a key process to qualify sustainable development. Vertical changes need to be assessed, and various methods for 3D change detection have been published. However, there is no common quantitative benchmark assessing their performance in urban areas yet. In this paper, we aim to fill this gap and introduce a simulation tool to generate synthetic 3D point cloud data in a well-controlled scenario. These data are then used to compare qualitatively and quantitatively representative 3D change detection methods for urban areas. These methods are based on distance computation (DSMd, C2C, M3C2), traditional machine learning (RF with stability feature) and deep learning (Feed Forward and Siamese networks). We distinguish between binary and multi-class classification of changes at different levels (3D points, 2D pixels, and 2D patches). While deep neural networks have led to numerous success in remote sensing, we show that they do not systematically outperform more simple methods for 3D change detection. Besides, the existing networks are limited to 2D patches while outputs at the pixel or point scale are more attractive.

Topo-bathymetric lidar is a powerful tool to survey coastal ecosystems while ensuring data continuity between land and water regardless of the nature of the terrain, and allowing the collection of information up to several dozens of metres deep. This study analyzes the potential of full waveform lidar data to monitor key ecosystems for climate change mitigation: seagrasses. It proposes an original way of processing topo-bathymetric lidar waveforms to map their spatial repartition and extent in Corsica (France). Waveform statistical and shape parameters are computed and used to produce a map of seagrass meadows that reaches over 86% of overall accuracy. Seagrass height is also extracted, offering perspectives for structural complexity assessment and ecosystem services quantification.

LiDAR point clouds are receiving a growing interest in remote sensing as they provide rich information to be used independently or together with optical data sources, such as aerial imagery. However, their nonstructured and sparse nature make them difficult to handle, conversely to raw imagery for which many efficient tools are available. To overcome this specific nature of LiDAR point clouds, the standard approach relies on converting the point cloud into a digital elevation model, represented as a 2-D raster. Such a raster can then be used similarly as optical images, e.g., with 2-D convolutional neural networks (CNNs) for semantic segmentation. In this letter, we show that LiDAR point clouds provide more information than only the digital elevation model and that considering alternative rasterization strategies helps to achieve better semantic segmentation results. We illustrate our findings on the IEEE Data Fusion Contest (DFC) 2018 data set.

This paper deals with strategies for LiDAR data analysis. While a large majority of studies first rasterize 3D point clouds onto regular 2D grids and then use 2D image processing tools for characterizing data, our work rather suggests to keep as long as possible the 3D structure by computing features on 3D data and rasterize later in the process. By this way, the vertical component is still taken into account. In practice, a voxelization step of raw data is performed in order to exploit mathematical tools defined on regular volumes. More precisely, we focus on attribute profiles that have been shown to be very efficient features to characterize remote sensing scenes. They require the computation of an underlying hierarchical structure (through a Max-Tree). Experimental results obtained on urban LiDAR data classification support the performances of this strategy compared with an early rasterization process.

Cloud-contaminated pixels exist ubiquitously in satellite images, which limit the usability of satellite images and increase the difficulty of image analysis. To reconstruct these pixels, a basic idea is to transfer cloud-free pixels from corresponding multi-temporal images to the target image, and the performance of this category of methods depends on the quality of information transfer between images. We propose in this work a novel pixel reconstruction method based on optimal transport. Our method first conducts an adaptive col-or transfer between multi-temporal images and then replaces cloud-contaminated pixels by transferred cloud-free pixels. The proposed method fully explores the potential of optimal transport to generate a more adaptive color transfer plan and thus ensure a high quality information transfer between images. Compared with other widely used methods, visual and statistical results on Landsat and MODIS images demonstrate the capacity of our method.

Mapping vegetation in the tropics is of primary importance to assess its contribution to important ecosystem services. This implies to implement methods to capture the vegetation gradient that characterizes land cover in these regions. Linear Mixture Models have long been used to monitor this gradient. In the present study, we automatically unmixed six Landsat 8 images of a study area in the Republic of Congo. We then computed the weighted average fraction of mineral, vegetation and water/shadow classes for each pixel in order to produce an annual (nearly) cloud-free unmixed image. Finally this product is pre-classified into six semantic classes ranging from “very dark” to “very bright” classes to discriminate the vegetation gradient based on its visual appearance. Results indicate the ability of the approach to classify fine land cover classes while still keeping textural information of the raw image.

One of the main objective of data fusion is the integration of several acquisition of the same physical object, in order to build a new consistent representation that embeds all the information from the different modalities. In this paper, we propose the use of optimal transport theory as a powerful mean of establishing correspondences between the modalities. After reviewing important properties and computational aspects, we showcase its application to three remote sensing fusion problems: domain adaptation, time series averaging and change detection in LIDAR data.

Satellite image time series (SITS) analysis is attracting more researchers recently because SITS have the advantage of fully capturing the dynamic changes of land cover and SITS data is becoming increasingly available. As an unsupervised classification method, clustering gains more importance due to frequent updates of labeled data or training samples are too expensive. When discussing SITS clustering, most researches focus on the similarity measure rather than clustering algorithm. However, the drawbacks of currently popular clustering algorithms tend to be amplified when tackling with SITS datasets. Therefore in this paper, we focus on the clustering algorithms and we find a novel method called affinity propagation is more suitable for SITS clustering. To demonstrate the accuracy and other advantages of affinity propagation, we conduct clustering experiments on MODIS and Landsat-TM SITS datasets. The obtained clustering maps are evaluated both visually and statistically comparing with other widely used clustering algorithms.

Hyperspectral (HS) images are commonly used in the context of planetary exploration, particularly for the analysis of the composition of planets. As several instruments have been sent throughout the Solar System, a huge quantity of data is getting available for the research community. Among classical problems in the analysis of HS images, a crucial one is unsupervised nonlinear spectral unmixing, which aims at estimating the spectral signatures of elementary materials and determining their relative contribution at a subpixel level. While the unmixing problem is well studied for Earth observation, some of the traditional problems encountered with Earth images are somehow magnified in planetary exploration. Among them, large image sizes, strong nonlinearities in the mixing (often different from those found in the Earth images), and the presence of anomalies are usually impairing the unmixing algorithms. This paper presents a new method that scales favorably with the problem posed by this analysis. It performs an unsupervised unmixing jointly with anomaly-detection capacities and has a global linear complexity. Nonlinearities are handled by decomposing the HS data on an overcomplete set of spectra, combined with a specific sparse projection, which guarantees the interpretability of the analysis. A theoretical study is proposed on synthetic data sets, and results are presented over the challenging 4-Vesta asteroid data set.

This paper presents a new method to estimate left ventricle deformations using variational data assimilation that combines image observations from cine MRI and a dynamic evolution model of the heart. The main contribution of the model is that it embeds parameters modeling the contraction / relaxation process. It estimates myocardial motion and contraction parameters simultaneously, providing accurate complementary information for diagnosis. The method was applied to synthetic datasets with known ground truth motion and to 47 patients MRI datasets acquired at three slice locations (base, mid-ventricle and apex). Radial and circumferential strain components were compared to those obtained with a reference tag tracking software, exhibiting good agreement with intraclass correlation coefficients (ICC) above 0.8. Results were also evaluated against wall motion score indices used to assess cardiac kinetics in clinical practice. The assimilation process overcame issues caused by temporal artifacts as a result of the dynamic model, compared to using the observation term alone. Moreover we found that the new dynamic model, consisting of a piecewise transport model acting independently on systole and diastole performed better than the standard continuous transport model, which oversmooths temporal variations. Estimated strain and contraction parameters significantly correlated to clinical scores, making them promising features for diagnosing not only hypokinesia but also dyskinesia.

The objective of this study was to evaluate how the combined use of multi-temporal optical and radar data can improve the precision of crop estimation in taking into account both discontinuous information on green vegetation and continuous information on vegetation cover.

In this paper, an original observation model for multiresolution optical flow estimation is introduced. Multiresolution frameworks, often based on coarse-to-fine warping strategies, are widely used by state-of-the-art optical flow methods. They allow the recovery of large motions by successive estimations of the flow field at several resolution levels. Although such approaches perform very efficiently and usually lead to faster minimizations, they generally consider independent problems at each resolution levels and do not exploit the existing interactions between scales (especially the influences of fine scales on larger ones). In this paper, we tackle this issue by proposing a flexible framework, inspired from fluid mechanics, able to partly counter these limitations. For each resolution level, our process filters the equations of interest and decomposes the key variables into resolved (i.e., at a given resolution) and unresolved (i.e., at finer resolutions) components. This enables to derive a new data term that takes into account, at coarse resolutions, the influence of their unresolved parts. From this new term, we propose two different estimation strategies, depending on whether we explicitly know the type of relations between the different scales (as for physical processes) or not. In order to test the efficiency of this new observation model, we have embedded it in a simple multiresolution Lucas-Kanade estimator. Comparing the usual optical flow constraint equation with this new term in the same motion estimation procedure, it clearly appears that the proposed term leads to more consistent estimates and prevents from errors propagation apparition during the estimation. In all situations (synthetic, real, physical images or not), our new term is able to greatly improve the results compared with usual conservation constraints.

Multi-scale frameworks based on coarse-to-fine warping strategies are widely used in the state-of-the-art optical flow methods. While they allow the estimation of large motions and usually lead to a faster minimization, they can also create strong dependencies between the successive estimates at different scale levels, yielding sometimes the propagation of undesirable errors from coarse to fine scales without any mean of correction. In this paper, we propose a more flexible framework inspired from fluid mechanics able to partly counter this issue. It relies on filtering equations where the variable of interest (i.e. the velocity field) is decomposed into resolved and unresolved components at each scale. We then derive a new data term that allows to take into account, in the coarse scales, information about smaller scale levels in order to avoid errors propagation during the estimation. Embedded in a simple Lucas-Kanade estimator, our new term is able to greatly improve the results from usual conservation constraints, as shown in the experimental part.

Grasslands, and more precisely agricultural practices associated with grasslands have an important impact on water and soil quality and biodiversity. In many regions, associated with agriculture intensification, a decrease of grasslands and change in their management can be observed during the last half century. Thus, determination of grassland management types represents an important approach for the quality and preservation of the environment. In this context, the objective of this study is to identify agricultural practices on grasslands from a time series of high spatial resolution images. Based on training samples, the classification of the LAI temporal profiles extracted from satellite images was performed using 1-standard classification technique as KNN and 2- an advanced ones using temporal kernels based on Dynamic Time Warping. Results show that use of an advanced classification technique improves of 20% the quality of grassland management identification.

In this paper, a stochastic formulation of the brightness consistency used in many computer vision problems involving dynamic scenes (for instance, motion estimation or point tracking) is proposed. Usually, this model, which assumes that the luminance of a point is constant along its trajectory, is expressed in a differential form through the total derivative of the luminance function. This differential equation linearly links the point velocity to the spatial and temporal gradients of the luminance function. However, when dealing with images, the available information only holds at discrete time and on a discrete grid. In this paper, we formalize the image luminance as a continuous function transported by a flow known only up to some uncertainties related to such a discretization process. Relying on stochastic calculus, we define a formulation of the luminance function preservation in which these uncertainties are taken into account. From such a framework, it can be shown that the usual deterministic optical flow constraint equation corresponds to our stochastic evolution under some strong constraints. These constraints can be relaxed by imposing a weaker temporal assumption on the luminance function and also in introducing anisotropic intensity-based uncertainties. We also show that these uncertainties can be computed at each point of the image grid from the image data and hence provide meaningful information on the reliability of the motion estimates. To demonstrate the benefit of such a stochastic formulation of the brightness consistency assumption, we have considered a local least-squares motion estimator relying on this new constraint. This new motion estimator significantly improves the quality of the results.

In this paper we address the joint interpolation of missing data and estimation of ocean surface velocities from multi-resolution sea surface satellite observations. A variational assimilation model is proposed. Using synthetic simulation and real SST data, we conducted experiments to evaluate the relevance of the proposed model, in particular the relevance of the fusion of observations at different resolutions and the improvement issued from the consideration of dynamics prior in the assimilation model. Numerical and qualitative results assessed the effectiveness of the proposed methods.

This paper is concerned with “structural” change detection in pair of images. This is a challenging and open problem since the difficulties stemming from the confusion between real changes (depending on the objects/structures inside the images) and visual changes (observed through the difference in terms of image luminance) are numerous. We propose to solve this labeling problem as the minimization of a global cost-function using a min-cut/max-flow strategy. Because of the different nature of the input images (different sensor, shooting angle, ...) and of the variety of detailed information contained in an object, we propose to rely on several criteria, either able to detect abrupt or subtle changes. These criteria are computed on local patches whose size adaptively depend on the structure of the objects inside the images. Experimental quantitative and qualitative results are shown on synthetic and real data.

This paper is concerned with "structural" change detection in pair of very high resolution remote sensing images. This is a challenging and open problem since the difficulties stemming from the confusion between real changes (depending on the objects/structures inside the images) and visual changes (observed through the difference in terms of image luminance) are numerous. Many applications are concerned with this crucial task (agriculture, urban,...). We propose to solve this labeling problem as the minimization of a global cost-function using a min-cut/max-flow strategy. Because of the different nature of the input images (different sensor, shooting angle, ...) and of the variety of detailed information contained in an object, we propose to rely on several criteria, either able to detect abrupt or subtle changes. These criteria are computed on local patches whose size adaptively depend on the structure of the objects inside the images. Experimental quantitative and qualitative results are shown on synthetic and real data.

This paper presents static and dynamic variational data assimilation methods for missing data interpolation in sea surface temperatures (SST) images. Evaluation using 50 AVHRR METOP SST images assesses the effectiveness of the proposed methods.

In this paper, a methodology for the spatial identification and characterization of vineyards using texture analysis is proposed to meet the need of ongoing and further viticultural “terroir” studies. The proposed method is based on the maximization of a criteria that deals with the coefficients enclosed in the different bands of a wavelet decomposition of the original image. More precisely, we search for the orientation that best concentrates the energy of the coefficients in a single direction. For each texture pattern, a degree of anisotropy and the angle of the main orientation is extracted. The methodology is validated on aerial-photographs in the Helderberg Basin (South Africa). The degree of anisotropy is a reliable information able to discriminate vineyards to other land-uses. Moreover, the row orientation turns out to be a relevant information for all applications related to mesoscale atmospheric modeling in vineyard areas.

This paper focuses on the tracking and analysis of convective cloud systems from Meteosat Second Generation images. The highly deformable nature of convective clouds, the complexity of the physical processes involved, and also the partially hidden measurements available from image data make difficult the direct use of conventional image-analysis techniques for tasks of detection, tracking, and characterization. In this paper, we face these issues using variational-data-assimilation tools. Such techniques enable us to perform the estimation of an unknown state function according to a given dynamical model and to noisy and incomplete measurements. The system state we are setting in this study for the cloud representation is composed of two nested curves corresponding to the exterior frontiers of the clouds and to the interior coldest parts (core) of the convective clouds. Since no reliable simple dynamical model exists for such phenomena at the image grid scale, the dynamics on which we are relying has been directly defined from image-based motion measurements and takes into account an uncertainty modeling of the curve dynamics along time. In addition to this assimilation technique, we show in the Appendix how each cell of the recovered cloud system can be labeled and associated to characteristic parameters (birth or death time, mean temperature, velocity, growth, etc.) of great interest for meteorologists.

This paper address the problem of change detection in very high resolution remote sensing images. To that end, we define a measure of the observed change based on the distribution of the coefficients issued from a wavelet transform, taking care to be rotation invariant. The dissimilarities are obtained through the Kullback-Liebler distance and a change features vector is defined from all the distances between the bands of the wavelet decomposition. This measurement is able to classify the nature of the change between two images. We present two applications: the first one uses a decision tree to classify several changes (homogeneous or oriented texture, abrupt or subtle change) whereas the second one detects some particular changes from a pair of images (an aerial and a satellite image). These experiments bring out the efficiency of the proposed technique to discriminate correctly the different textures and to interpret each change.

This paper is interested with the tracking and the analysis of convective cells on Meteosat Second Generation images. Due to the highly deformable nature of convective clouds, the noisy measurements obtained with image data and the complexity of the physics that governs such phenomena, the conventional tools issued from computer vision to detect and track usual (rigid) objects are not well adapt. In this paper, we face this problem using variational data assimilation tools. Such techniques enable to perform the estimation of an unknown state function according to a given dynamical model and to noisy and incomplete measurements. The system state is composed of two curves (represented with implicit surfaces) corresponding to the the whole cloud and the coldest part (heart) of the convective system. Since no precise and manageable dynamical model exists concerning such phenomena, the dynamics is directly measured from images using some motion estimation techniques devoted to fluid motions.

The complexity of dynamical laws governing 3D atmospheric flows associated with incomplete and noisy observations make the recovery of atmospheric dynamics from satellite images sequences very difficult. In this paper, we face the challenging problem of estimating physical sound and time-consistent horizontal motion fields at various atmospheric depths for a whole image sequence. Based on a vertical decomposition of the atmosphere, we propose a dynamically consistent atmospheric motion estimator relying on a multi-layer dynamical model. This estimator is based on a weak constraint variational data assimilation scheme and is applied on noisy and incomplete pressure difference observations derived from satellite images. The dynamical model consists in a simplified vorticity-divergence form of a multi-layer shallow-water model. Average horizontal motion fields are estimated for each layer. The performance of the proposed technique is assessed on real world meteorological satellite image sequences.

The objective of this study is to identify changes in the vegetation cover in estimating crop fraction cover in the Brittany region over the 2000-2006 period. Biophysical products derived from current coarse and medium resolution sensors allow the detection of changes over landscapes made of large patches. In fragmented landscapes, the 1 km spatial resolution is not suited to identify changes in vegetation cover due to mixing effects. Higher spatial resolution is thus required to detect vegetation changes. We present an original method to estimate biophysical variables and particularly fCOVER at 250 m spatial resolution from EOS/MODIS-AM1 (Terra) sensor in using the combined PROSPECT-SAIL model The results obtained from this study highlight that MODIS fCOVER data at 250 m spatial resolution estimated with the PROSPECT+ SAIL model is related to field observations of green vegetation cover fraction.

The objective of this paper is to develop an object-based change detection method able to qualify the nature of changes in landscapes from remotely sensed images, in terms of geometry and content. The originality of the approach consists in jointly dealing with the analysis of the object contours and the analysis of texture evolution. The method is applied on grassy strips, which are landscape buffers between crops and hydrologic networks. A geometric change index that quantify the intensity of change and that qualify it according its properties is proposed. We also present a content change index that discriminates partial change from diffuse change that affect the object of interest. It turned out that the geometric changes, most of abrupt content changes and some of subtle changes have been accurately detected. Lastly, our approach is suitable on airborne data with a Very High Resolution data and can be generalized to spaceborne images.

In this paper, we present a framework for dynamic consistent estimation of dense motion fields over a sequence of images. The originality of the approach is to exploit recipes related to optimal control theory. This setup allows performing the estimation of an unknown state function according to a given dynamical model and to noisy and incomplete measurements. The overall process is formalized through the minimization of a global spatio-temporal cost functional w.r.t the complete sequence of motion fields. The minimization is handled considering an adjoint formulation. The resulting scheme consists in iterating a forward integration of the evolution model and a backward integration of the adjoint evolution model guided by a discrepancy measurement between the state variable and the available noisy observations. Such an approach allows us to cope with several delicate situations (such as the absence of data) which are not well managed with usual estimators.

In this paper, we present a framework for dynamic consistent estimation of dense motion fields over a sequence of Meteosat Second Generation (MSG) images. The originality of the approach is to exploit recipes related to optimal control theory developed in geophysical sciences. This framework, known as variational data assimilation, enables to perform the estimation of an unknown state function according to a given dynamical model and to noisy and incomplete measurements. In our work, the measurements are defined according to a smoothed brightness consistency model whereas the dynamical model on which we rely is derived from a velocity conservation law. The overall assimilation process is formalized through the minimization of a global spatio- temporal cost functional w.r.t to the complete sequence of motion fields. The minimization is handled considering an adjoint formulation. The resulting scheme consists in iterating a forward integration of the evolution model and a backward integration of the adjoint evolution model guided by a discrepancy measurement between the state variable and the available noisy observations. Such an approach allows us to cope with several delicate situations (such as the absence of data) which are not well managed with usual estimators. The efficiency of our approach is demonstrated on real data. It enables to estimate a sequence of dense motion fields even in situations where data are strongly corrupted.

Abstract In this paper a new and original technique to animate a crowd of human beings is presented. Following the success of data‐driven animation models (such as motion capture) in the context of articulated figures control, we propose to derivate a similar type of approach for crowd motions. In our framework, the motion of the crowds are represented as a time series of velocity fields estimated from a video of a real crowd. This time series is used as an input of a simple animation model that ‘advect’ people along this time‐varying flow. We demonstrate the power of our technique on both synthetic and real examples of crowd videos. We also introduce the notions of crowd motion editing and present possible extensions to our work. Copyright © 2007 John Wiley & Sons, Ltd.

In this paper, we present an active contour method able to capture texture boundaries. To that end, an external potential term based on wavelet coefficients is defined. The proposed approach only deals with boundaries to be able to detect particular patterns which are identifiable only through a texture border. Such patterns, which cannot be treated like traditional objects, appear in many natural images (meteorology, biology or medical imaging). Their presence often inform, about a particular phenomena in the observed scene and the detection of such structures is hence of primary interest. The different bands of wavelet coefficients enable to characterize the desired texture, They also correctly deal with the delicate situations of transparency or partial occlusions.

International audience

In this paper, we address the problem of estimating and analyzing the motion of fluids in image sequences. Due to the great deal of spatial and temporal distortions that intensity patterns exhibit in images of fluids, the standard techniques from computer vision, originally designed for quasi-rigid motions with stable salient features, are not well adapted in this context. We thus investigate a dedicated minimization-based motion estimator. The cost function to be minimized includes a novel data term relying on an integrated version of the continuity equation of fluid mechanics, which is compatible with large displacements. This term is associated with an original second-order div-curl regularization which prevents the washing out of the salient vorticity and divergence structures. The performance of the resulting fluid flow estimator is demonstrated on meteorological satellite images. In addition, we show how the sequences of dense motion fields we estimate can be reliably used to reconstruct trajectories and to extract the regions of high vorticity and divergence.

We address the problem of fluid motion estimation in image sequences. For such motions, standard optical flow methods, based on intensity conservation and spatial coherence of motion field, are not suitable. This is due to the highly deformable nature of a fluid medium. For all applications where fluid motions are to be recovered from images, it is then important to have specific techniques. We investigate such dedicated models which include an original observation constraint, based on the continuity equation from fluid mechanics, and a new div-curl-type smoothness term. Our method is validated on synthetic and real meteorological images.