• Refine Query
  • Source
  • Publication year
  • to
  • Language
  • 1
  • Tagged with
  • 3
  • 3
  • 2
  • 2
  • 2
  • 1
  • 1
  • 1
  • 1
  • 1
  • 1
  • 1
  • 1
  • 1
  • 1
  • About
  • The Global ETD Search service is a free service for researchers to find electronic theses and dissertations. This service is provided by the Networked Digital Library of Theses and Dissertations.
    Our metadata is collected from universities around the world. If you manage a university/consortium/country archive and want to be added, details can be found on the NDLTD website.
1

Restauration d'images par temps de brouillard et de pluie : applications aux aides à la conduite / Image restoration during foggy and rainy weather : applications for driver assistance systems

Halmaoui, Houssam 30 November 2012 (has links)
Les systèmes d'aide à la conduite (ADAS) ont pour objectif d'assister le conducteur et en particulier d'améliorer la sécurité routière. Pour cela, différents capteurs sont généralement embarqués dans les véhicules afin, par exemple, d'avertir le conducteur en cas de danger présent sur la route. L'utilisation de capteurs de type caméra est une solution économiquement avantageuse et de nombreux ADAS à base de caméra voient le jour. Malheureusement, les performances de tels systèmes se dégradent en présence de conditions météorologiques défavorables, notamment en présence de brouillard ou de pluie, ce qui obligerait à les désactiver temporairement par crainte de résultats erronés. Hors, c'est précisément dans ces conditions difficiles que le conducteur aurait potentiellement le plus besoin d'être assisté. Une fois les conditions météorologiques détectées et caractérisées par vision embarquée, nous proposons dans cette thèse de restaurer l'image dégradée à la sortie du capteur afin de fournir aux ADAS un signal de meilleure qualité et donc d'étendre la gamme de fonctionnement de ces systèmes. Dans l'état de l'art, il existe plusieurs approches traitant la restauration d'images, parmi lesquelles certaines sont dédiées à nos problématiques de brouillard ou de pluie, et d'autres sont plus générales : débruitage, rehaussement du contraste ou de la couleur, "inpainting"... Nous proposons dans cette thèse de combiner les deux familles d'approches. Dans le cas du brouillard notre contribution est de tirer profit de deux types d'approches (physique et signal) afin de proposer une nouvelle méthode automatique et adaptée au cas d'images routières. Nous avons évalué notre méthode à l'aide de critères ad hoc (courbes ROC, MSE, contraste visibles à 5 %, évaluation sur ADAS) appliqués sur des bases de données d'images de synthèse et réelles. Dans le cas de la pluie, une fois les gouttes présentes sur le pare-brise détectées, nous reconstituons les parties masquées de l'image à l'aide d'une méthode d'"inpainting" fondée sur les équations aux dérivées partielles. Les paramètres de la méthode ont été optimisés sur des images routières. Enfin, nous montrons qu'il est possible grâce à cette approche de construire trois types d'applications : prétraitement, traitement et assistance. Dans chaque famille, nous avons proposé et évalué une application spécifique : détection des panneaux dans le brouillard ; détection de l'espace navigable dans le brouillard ; affichage de l'image restaurée au conducteur. / Advanced Driver Assistance Systems (ADAS) are designed to assist the driver and in particular to improve road safety. For this purpose, various sensors are typically embedded in vehicles in order, for example, to alert the driver in case of imminent danger on the road. The use of camera type of sensor is a cost-effective solution and many ADAS based on camera are being created. Unfortunately, the performance of such systems decrease drastically in the presence of adverse weather conditions, especially in the presence of fog or rain, which could oblige to turn off the systems temporarily in order to avoid erroneous results. While, it is precisely in these difficult circumstances that the driver would potentially need the most to be assisted. Once the weather conditions detected and characterized by embedded vision, we propose in this thesis to restore the degraded image to provide a better signal to the ADAS and thus extend the operation range of these systems. In the state of the art, there are several approaches dealing with images restoration, some of which are dedicated to our fog and rain problem and others are more general : denoising, contrast or color enhancement, inpainting... We propose in this work to combine the two families of approaches. In the case of fog our contribution is to take advantage of both approaches (physical and signal) to propose a new automatic method adapted to the case of road images. We evaluated our method using ad hoc criteria (ROC curves, visible contrast to 5%, assessment on ADAS) applied to databases of synthetic and real images. In case of rain, once the drops present on the windshield are detected, we reconstruct the hidden parts of the image using a method of inpainting based on partial differential equations. The method parameters have been optimized on road images. Finally, we show that it is possible with this approach to build three types of applications : preprocessing, processing and assistance. In every family, we have proposed and evaluated a specific application : traffic signs detection during foggy weather; detection of free space in fog conditions and display of the restored image to the driver.
2

Ocean Rain Detection and Wind Retrieval Through Deep Learning Architectures on Advanced Scatterometer Data

McKinney, Matthew Yoshinori Otani 18 June 2024 (has links) (PDF)
The Advanced Scatterometer (ASCAT) is a satellite-based remote sensing instrument designed for measuring wind speed and direction over the Earth's oceans. This thesis aims to expand and improve the capabilities of ASCAT by adding rain detection and advancing wind retrieval. Additionally, this expansion to ASCAT serves as evidence of Artificial Intelligence (AI) techniques learning both novel and traditional methods in remote sensing. I apply semantic segmentation to ASCAT measurements to detect rain over the oceans, enhancing capabilities to monitor global precipitation. I use two common neural network architectures and train them on measurements from the Tropical Rainfall Measuring Mission (TRMM) collocated with ASCAT measurements. I apply the same semantic segmentation techniques on wind retrieval in order to create a machine learning model that acts as an inverse Geophysical Model Function (GMF). I use three common neural network architectures and train the models on ASCAT data collocated with European Centre for Medium-Range Weather Forecasts (ECMWF) wind vector data. I successfully increase the capabilities of the ASCAT satellite to detect rainfall in Earth's oceans, with the ability to retrieve wind vectors without a GMF or Maximum Likelihood Estimation (MLE).
3

An Analysis of SeaWinds Simultaneous Wind/Rain Retrieval in Severe Weather Events

Allen, Jeffrey R. 08 March 2005 (has links) (PDF)
Scatterometers, such as SeaWinds, can provide wide coverage of ocean surface winds. They estimate near-surface wind vectors by relating measured radar backscatter to a geophysical model function. However, SeaWinds measurements are also sensitive to rain, and conventional wind retrieval degrades in rainy conditions. An algorithm that exploits SeaWinds' sensitivity to both wind and rain has be developed. This algorithm, termed simultaneous wind/rain retrieval, retrieves both wind vectors and rain rates for a given ocean area. Instantaneous results of simultaneous wind/rain retrieval in Hurricane events is analyzed through comparison with the NEXRAD ground-based radar system. This comparison allows validation of retrieved rains. Additionally, conditions that affect the accuracy of SeaWinds wind/rain observations are evaluated. It is shown that, when thresholded, the rains retrieved by SeaWinds give an adequate rain flag. The comparisons of SeaWinds and NEXRAD rain estimates facilitate construction of a model to simulate variability in the SeaWinds rain estimates. The model is used to show that rain estimates are unbiased, though with significant variability. The variability is likely to be primarily driven by the noise inherent to the SeaWinds system.

Page generated in 0.1017 seconds