Letteratura scientifica selezionata sul tema "Imagerie passive – Innovation"

Abstract. This study assesses the potential of 2 to 10 km resolution imagery of CO2 concentrations retrieved from the shortwave infrared measurements of a space-borne passive spectrometer for monitoring the spatially integrated emissions from the Paris area. Such imagery could be provided by missions similar to CarbonSat, which was studied as a candidate Earth Explorer 8 mission by the European Space Agency (ESA). This assessment is based on observing system simulation experiments (OSSEs) with an atmospheric inversion approach at city scale. The inversion system solves for hourly city CO2 emissions and natural fluxes, or for these fluxes per main anthropogenic sector or ecosystem, during the 6 h before a given satellite overpass. These 6 h correspond to the period during which emissions produce CO2 plumes that can be identified on the image from this overpass. The statistical framework of the inversion accounts for the existence of some prior knowledge with 50 % uncertainty on the hourly or sectorial emissions, and with ∼ 25 % uncertainty on the 6 h mean emissions, from an inventory based on energy use and carbon fuel consumption statistics. The link between the hourly or sectorial emissions and the vertically integrated column of CO2 observed by the satellite is simulated using a coupled flux and atmospheric transport model. This coupled model is built with the information on the spatial and temporal distribution of emissions from the emission inventory produced by the local air-quality agency (Airparif) and a 2 km horizontal resolution atmospheric transport model. Tests are conducted for different realistic simulations of the spatial coverage, resolution, precision and accuracy of the imagery from sun-synchronous polar-orbiting missions, corresponding to the specifications of CarbonSat and Sentinel-5 or extrapolated from these specifications. First, OSSEs are conducted with a rather optimistic configuration in which the inversion system is perfectly informed about the statistics of the limited number of error sources. These OSSEs indicate that the image resolution has to be finer than 4 km to decrease the uncertainty in the 6 h mean emissions by more than 50 %. More complex experiments assess the impact of more realistic error estimates that current inversion methods do not properly account for, in particular, the systematic measurement errors with spatially correlated patterns. These experiments highlight the difficulty to improve current knowledge on CO2 emissions for urban areas like Paris with CO2 observations from satellites, and call for more technological innovations in the remote sensing of vertically integrated columns of CO2 and in the inversion systems that exploit it.

Environmental interactions of marine renewable energy (MRE) projects are challenging to monitor, and key questions remain about their potential impacts. Processing large volumes of environmental data acquired from submarine monitoring and the use of machine learning to identify presence and interactions of marine wildlife with MRE infrastructure are powerful tools for assessing the environmental response to MRE infrastructure. The use of automated image analysis for species identification and enumeration using algorithms like convolutional neural networks can vastly reduce the time required to extract usable data from submarine imagery compared to manual expert processing. We present a novel industry-ready image processing workflow for automated wildlife detection developed using 1000+ hours of underwater video footage obtained by Nova Innovation Ltd. from their operational tidal stream turbine array at Bluemull Sound in Shetland, Scotland. The objective of this work was to develop a workflow and associated algorithms to automatically filter many hours of underwater video, remove unwanted footage, and extract only video containing marine mammals, diving birds or fish. The workflow includes object detection through advanced image analysis, image classification using machine learning, statistical analyses such as quantification of data storage reduction and number of detections, and automated production of a summary report. Blind tests were undertaken on a subset of videos to quantify and iteratively improve the accuracy of the results. The final iteration of the workflow delivered an accuracy of 80% for the identification of marine mammals, diving birds and fish when a three-category (wildlife, algae, and background) classification system was used. The accuracy rose to 95% when a two-category system was used, and objects were classified simply as ‘target’ or ‘non-target’. The entire workflow can be run from video inception to production of an automated results report in approximately 30 minutes, dependent on size of input data, when environmental conditions such as water clarity and key species of interest are familiar to the algorithm. The accuracy and runtime speed of the workflow can be improved through expanding the training dataset of images used in the development of this initial tool by including additional species and water conditions. Application of this workflow significantly reduces manual processing and interpretation time, which can be a significant burden on project developers. Automated processing provides a subset for more focused manual scrutiny and analysis, while reducing the overall size of dataset requiring storage. Auto-reporting can be used to provide outputs for marine regulators to meet monitoring reporting conditions of project licences. Integration of this workflow with automated passive acoustic monitoring systems can provide a holistic environmental monitoring approach using both underwater imagery and acoustics.

L’imagerie passive s’appuie sur des algorithmes de formation de voie qui nécessitent des sondes de grande ouverture pour offrir de bonnes résolutions axiales ; or, en imagerie passive 3D, les sondes matricielles actuellement commercialisées ne vérifient pas cette contrainte. De plus, ces sondes présentent un grand nombre d’éléments, ce qui rend leur utilisation particulièrement lourde. Ce travail de thèse porte sur l’étude et l’amélioration de l’imagerie passive de la cavitation en s’intéressant à deux aspects en particulier : (i) la mise en œuvre pratique et efficace de l’imagerie passive en 3D, (ii) la problématique de l’imagerie de sources étendues telles que des nuages de cavitation. Nous avons combiné l’application des méthodes sparse (pour réduire le nombre d’éléments actifs de la sonde utilisée) et la transposition du 2D au 3D des algorithmes adaptatifs dans le domaine fréquentiel. Ce formalisme utilise l’estimation robuste de la matrice de densité inter-spectrale (CSM) et nous a permis d’implémenter simplement et efficacement différents algorithmes : Delay-And-Sum (DAS), Robust-Capon-Beamformer et Pisarenko. L’efficacité de ces algorithmes en 3D a été testée en termes de largeur à mi-hauteur, de contraste et d’erreur de position, sur une source ponctuelle en simulations et sur une expérience de réflecteur ponctuel. Enfin, pour répondre à la réalité des nuages de cavitation, nous nous sommes intéressés au comportement de ces méthodes de reconstruction dans le cas de sources étendues. Nos simulations en 2D montrent l’évolution des images reconstruites en fonction de caractéristiques du nuage de cavitation. Ce travail apporte une solution concrète de mise en œuvre simple de l’imagerie passive 3D ainsi que des éléments de réponse quant aux attentes sur la localisation et la caractérisation d’un nuage de cavitation
Passive imaging relies on beamforming algorithms that require large aperture probes to provide good axial resolutions; however, in 3D passive imaging, the matrix probes currently marketed do not meet this constraint. Moreover, these probes have a large number of elements, which makes their use particularly unwieldy. This thesis work focuses on the study and improvement of passive cavitation imaging by addressing two aspects in particular: (i) the practical and efficient implementation of 3D passive imaging, (ii) the problem of imaging large sources such as cavitation clouds. We have combined the application of sparse methods (to reduce the number of active elements of the probe used) and the transposition from 2D to 3D of adaptive algorithms in the frequency domain. This formalism uses the robust estimation of the inter-spectral density matrix (CSM) and allowed us to implement simply and efficiently different algorithms: Delay-And-Sum (DAS), Robust-Capon-Beamformer and Pisarenko. The efficiency of these algorithms in 3D has been tested in terms of width to half height, contrast and position error, on a point source in simulations and on a point reflector in experiments. Finally, in order to address the reality of cavitation clouds, we have investigated the behavior of these reconstruction methods in the case of extended sources. Our 2D simulations show the evolution of the reconstructed images as a function of the cavitation cloud characteristics. This work provides a concrete solution for a simple implementation of 3D passive imaging as well as answers to the expectations on the localization and characterization of a cavitation cloud

L'élastographie ondulatoire est une méthode d'imagerie visant à mesurer l'élasticité des tissus biologiques de façon non-invasive et quantitative. Récemment, la transposition de la technique à petite échelle baptisée micro-élastographie dynamique a permis de réaliser de premières mesures d'élasticité cellulaire par ondes de cisaillement grâce à un microscope optique. Cette thèse s'attache à en comprendre les limites et à développer de nouvelles méthodes de micro-élastographie, à tester de nouvelles sources d'ondes mais également des applications potentielles de la technique. Dans un premier temps, la dispersion d'ondes de cisaillement a été étudiée sur des gels de gélatine. Deux régimes distincts d'ondes élastiques guidées et d'ondes de cisaillement ont été identifiés. La limite haute fréquence de propagation des ondes a également été explorée, permettant d'établir l'existence d'une fréquence de coupure expliquant l'absence d'imagerie ultrasonore de cisaillement. La même approche a ensuite été appliquée à des fluides visco-élastiques faisant apparaître deux fréquences de coupure et permettant de revisiter les études déjà menées sur la rhéologie et la propagation d'ondes dans ce type de milieux. Puis, l'objectif initial étant de réaliser de la micro-élastographie sur des cellules uniques et les expériences précédemment réalisées avec des micro-pipettes présentant certains défauts, une méthode originale de micro-élastographie cellulaire a été développée. Une micro-bulle oscillante est utilisée comme source d'ondes de cisaillement sans contact à 15 kHz, pour réaliser des expériences sur des cellules sanguines appelées mégacaryocytes dont le diamètre est d'environ 15 µm. Il s'agit en fait des plus petits objets jamais explorés par élastographie. Des objets plus gros, des amas cellulaires de quelques dizaines de milliers de cellules ont également été étudiés. En effet, l'élastographie ultrasonore de ces modèles tumoraux d'environ 800 µm de diamètre étant impossible, la micro-élastographie optique est une technique adaptée. Ces échantillons contiennent des nano-particules magnétiques, donc une impulsion magnétique a pu être utilisée comme source d'ondes. Auparavant, des preuves de concept sur des gels à la fois macroscopiques (en élastographie ultrasonore) et microscopiques (en micro-élastographie optique) ont été menées pour valider l'utilisation de cette source de champ diffus. Enfin, des mesures d'ondes de pouls ont été réalisées sur des artères rétiniennes d'environ 50 µm de diamètre à partir d'acquisitions d'holographie Doppler laser réalisées in vivo. L'application d'algorithmes de corrélation monochromatiques a permis de mesurer la vitesse d'ondes guidées révélant l'existence d'une deuxième onde de pouls, une onde antisymétrique de flexion. Cette onde guidée, beaucoup plus lente que l'onde de pouls axisymétrique étudiée jusqu'à présent, a également été observée sur l'artère carotide grâce à des acquisitions ultrasonores ultrarapides
Dyanmic elastography is an imaging method to measure the elasticity of biological tissues in a non-invasive and quantitative way. Recently, the transposition of the technique to a small scale has been called dynamic micro-elastography and has allowed the first measurements of cellular elasticity by shear waves using an optical microscope. This thesis aims to undetstand the limits of this technique and to develop new micro-elastography methods, to test new wave sources but also potential applications of the technique. In a first step, the dispersion of shear waves was studied on gelatin phantoms. Two distinct regimes of guided elastic waves and shear waves were identified. The high-frequency limit of wave propagation was also explored, establishing the existence of a cutoff frequency which explains the absence of ultrasonic shear imaging. The same approach was then applied to visco-elastic fluids, revealing two cutoff frequencies and revisiting previous studies on rheology and wave propagation in this type of medium. Then, the initial objective being to carry out micro-elastography on single cells and the experiments previously carried out with micro-pipettes presenting certain defects, an original method of cellular micro-elastography was developed. An oscillating microbubble is used as a contactless shear wave source at 15 kHz to perform experiments on blood cells whose diameter is about 15 µm. These are the smallest objects ever explored by elastography. Larger objects, cell clusters of a few tens of thousands of cells have also been studied. Indeed, since ultrasound elastography of these tumour models of about 800 µm in diameter is impossible, optical micro-elastography is a suitable technique. These samples contain magnetic nanoparticles, so a magnetic pulse could be used as a wave source. Previously, proofs of concept on both macroscopic (in ultrasonic elastography) and microscopic (in optical micro-elastography) phantoms were conducted to validate the use of this diffuse field source. Finally, pulse wave measurements were performed on retinal arteries of about 50 µm in diameter using laser Doppler holography acquisitions performed in vivo. The application of monochromatic correlation algorithms allowed the measurement of guided wave velocities, finally revealing the existence of a second pulse wave, an antisymmetric bending wave. This guided wave, much slower than the axisymmetric pulse wave studied so far, was also observed on the carotid artery thanks to ultrafast ultrasound acquisitions

Abstract The offshore wind industry is promoting developments in environmental sensing, machine learning, and artificial intelligence, to better detect the presence of marine and avian species. Environmental sensing technologies (e.g., radar, video and infra-red imagery, passive acoustics, and radio telemetry) have advanced where wildlife are reliably detected and tracked, aiding their protection by minimizing conflicts with ships, other users of the ocean space, and other stressors. Significant marine ecosystem data is collected daily offshore from a wide range of reputable sources. These disconnected sources represent, in aggregate, a trove of Domain Awareness (DA) data and if cohesively viewed, provide opportunity to better de-risk operations, protect wildlife, and avoid delays in real time. Taking care and effort to assimilate these (often disparate) data sources into common visualization platform(s) provides both more granular and macro-scale situational awareness, while advancing opportunities to apply predictive Artificial Intelligence (AI) to the data. This can result in the application of regional (or broad scale) predictions and understandings of species activities. As this data base of predictions and observations grow, additional decision making and management mitigations can be applied, such as alerting specific vessels to the presence of protected species or initiating tailored dynamic management areas (DMAs) at appropriate temporal or spatial scales. Deployment of sensors on technically advanced host platforms, including autonomous underwater vehicles, uncrewed surface vehicles, and metocean buoys, is occurring regularly. Equally prolific are strategies to collect, analyze, and display data from each sensor, resulting in myriad data dashboards, digital twins, and immersive visualization environments offered to offshore wind developers and regulators. While accelerating technological innovation, these numerous, and often single-focus approaches can hinder the delivery of a unified picture of the worksite or regional environment, limiting conservation value of these efforts and increasing environmental and scheduling project risks. This paper reviews some of the extant initiatives to deliver environmental data and provides a suite of best practices and recommendations for developing a DA capability or a common operating picture (COP) of developer's projects, as well as a regional view that covers multiple worksites. This work will assist developers and regulators to understand a realistic state of technical readiness and how to appropriately scope data products that support data fusion consistently across visualization platforms.