Kopoos Consulting (France)

Significant efforts have been directed towards adapting self-supervised multimodal learning for Earth observation applications. However, most current methods produce coarse patch-sized embeddings, limiting their effectiveness and integration with other modalities like LiDAR. To close this gap, we present DUNIA, an approach to learn pixel-sized embeddings through cross-modal alignment between images and full-waveform LiDAR data. As the model is trained in a contrastive manner, the embeddings can be directly leveraged in the context of a variety of environmental monitoring tasks in a zero-shot setting. In our experiments, we demonstrate the effectiveness of the embeddings for seven such tasks: canopy height mapping, fractional canopy cover, land cover mapping, tree species identification, plant area index, crop type classification, and per-pixel waveform-based vertical structure mapping. The results show that the embeddings, along with zero-shot classifiers, often outperform specialized supervised models, even in low-data regimes. In the fine-tuning setting, we show strong performances near or better than the state-of-the-art on five out of six tasks.

Abstract Road transportation accounts for up to 35% of carbon dioxide and 49% of nitrogen oxides emissions in the Paris region. However, estimates of city traffic patterns are often incomplete and of coarse spatio-temporal resolution, even where extensive networks of sensors exist. This study uses a machine learning approach to analyze data from 2086 magnetic road sensors across Paris, generating a detailed dataset of hourly traffic flow and road occupancy covering 6846 road segments from 2018 to 2022. Our model captures flow and occupancy with a symmetric mean absolute percentage error of 37% and 54% respectively, providing high-resolution insights into traffic patterns. These insights allow for the creation of a comprehensive map of hourly transportation patterns in Paris, offering a robust framework for assessing traffic variables for each significant road link in the city. The model’s ability to incorporate an emission factor based on the mean speed of the vehicle fleet, derived from flow and occupancy data, holds promise for developing a detailed CO2 and pollutant inventory. This methodology is not limited to Paris; it can be applied to other urban centers with similar data availability, highlighting its potential as a versatile tool for sustainable urban monitoring.

In this work, discrete elements methods (DEM) were used to simulate ballasted track lateral resistance problems by representing a granular material such as ballast as an assembly of rigid grains interacting through contact laws that contain micromechanical physics relevant for the material.Macro-mechanical properties, such as global resistance of the ballast bed, are then seen as a consequence of the collective dynamics of the assembly.A large number of lateral resistance tests simulations were run to find optimal value for frictional parameters so that experimentally measured values for various track configurations were satisfyingly found.The model was then used to explore the lateral track resistance of experimentally untested configurations, namely standard profiles and nonstandard ballast profile alike, for instance no shoulder, reinforced shoulder, "infinite" shoulder or walls.By varying the length and height of the shoulder, we could predict how much one can enhance the lateral resistance by adding more ballast on the side of the track, and what geometry is optimal for profile reinforcement in term of ballast quantity, while respecting security constraints toward lateral efforts due to potential rail buckling.

Time series of satellite images are now massively available thanks to the existence of several constellations of recurrent satellites. We propose a method for detecting and measuring the duration of changes on such series. This approach is intended to be generic and independent of the type of satellite used, whether band limited or multispectral. It is based on a global analysis of the sequence. The statistical detection method is applied to a residual sequence computed from backward and forward novelty filters applied to all images in the series. Significant changes are computed with a guarantee on their number of false alarms (NFA). To establish the efficiency of the method, we have created an open database of 28 sequences of 20 images acquired by the Sentinel-2 satellite, in different regions of the world. We obtain satisfactory results, which are consistent with the visual observations.

Methane emissions monitoring is essential to control methane pollution. In this paper, we propose an automatic practical methodology using time series to estimate the quantity of methane in a given plume using a multispectral satellite like Sentinel-2. Sentinel-2 proposes a low revisit time, a good spatial resolution and a low acquisition cost. Contrary to previous methods, the proposed approach does not require a manual selection of an optimal reference image. We compared its performance on an oil-and-gas site in Kazakhstan. This is the first step toward an automatic global monitoring system for methane plume detection and quantification with these satellites.

The Rational Polynomial Camera (RPC) model can be used to describe a variety of image acquisition systems in remote sensing, notably optical and Synthetic Aperture Radar (SAR) sensors. RPC functions relate 3D to 2D coordinates and vice versa, regardless of physical sensor specificities, which has made them an essential tool to harness satellite images in a generic way. This article describes a terrain-independent algorithm to accurately derive a RPC model from a set of 3D-2D point correspondences based on a regularized least squares fit. The performance of the method is assessed by varying the point correspondences and the size of the area that they cover. We test the algorithm on SAR and optical data, to derive RPCs from physical sensor models or from other RPC models after composition with corrective functions.

Monitoring oil stocks provides valuable insights on the balance between production and demand of petroleum products. The identification of oil depots is important for estimating storage capacities and measuring oil stocks. To achieve this purpose, we present an oil tank detector. As oil tanks are generally circular, we use as a first step a circle detection algorithm. However, this approach tends to generate a considerable amount of false detections, as circular shapes are not necessarily tanks. To reduce these false detections, we take advantage of the fact that oil tanks are generally densely grouped together, and filter out isolated detections using a clustering algorithm. The clustering method uses the a contrario framework which gives a way to control the number of false detections per image. The method is illustrated on Sentinel-2 images.

Abstract. The storage and management of stockpiles of materials is a fundamental process in large scale activities such as mining, civil engineering, and in the management of waste landfill sites. Following the evolution of stocks has always been important, and advancements in remote sensing technology are not only facilitating this, but also making it possible in near real-time. Nowadays, this monitoring appears to be performed almost exclusively using UAV based techniques. This paper proposes to apply a simple Shape-from-Shading method on low resolution satellite images provided by PlanetScope to monitor the evolution of the volume of stockpiles. The proposed Shape-from-Shading formulation makes it possible to handle occluding objects in the scene. The loss in accuracy due to the low resolution of PlanetScope images is well compensated by the daily revisit frequency and by the fact that spaceborne acquisitions require no human supervision. With satellites it is also easy to follow simultaneously several stocking sites all over the world. We test our method on two coal storage sites and demonstrate that the stockpiles are well detected.