<p>The first part of the thesis consists of a brief introduction to the general principles of target detection and the sensor-systems used. In the following part there is a theoretical description of the algorithms this thesis focuses on. The detection algorithms described in this paper are called Cell Average, Ordered Statistics, 2parameter and Gammadetector. Two different discriminators called Extended Fractal Features and Quadratic Gamma Discriminator are also described. The algorithms are tested on three different types of data, simulated SAR-pictures, authentic SAR-targets and IR-pictures. The last part account for the results, both those achieved with pictures and those results achieved when doing statistical tests, in this case MonteCarlo- simulations and Reciever Operating Characteristics-curves. The results show that the Gamma- detector and the QGD-algorithm perform best on the tests done in this thesis.</p>

The technique of ESR-dosimetry and strategies for investigation of new materials as in regard to their applicability as ESR-dosimeters for radiotherapy has been reviewed. As an example six salts of formic and lactic acid has been evaluated. The applicability of the dosimeter has been judged by evaluating the tissue equivalence, radical yield, radical stability, spectral suitability, optimal readout parameters, dose response and sensitivity of the dosimetric system. Dependence of material characteristics and influence parameters has been analysed. The reviewed methods have been successfully used for evaluation of the new materials. Lithium formate has been shown to be a good candidate relative to the state of the art dosimeter of alanine. Using optimal readout parameters lithium formate has been shown to be nine times as sensitive but even at moderate settings lithium formate is more sensitive. The results for lithium formate are in accordance to those of previous studies. The signal intensity of sodium formate has also proved to be high but unfortunately the signal fades rapidly. Two new methods have been proposed as synthesis of the reviewed methods. The first allows flexible, effective and objective baseline correction of the ESR-spectrum. The second deals with dose response measurement by linear regression of the entire spectrum and was found to be successful in separating the spectral peaks of the induced radicals from the background signal.

Nous nous sommes intéressés au problème de l'extraction d'information dans des images (Image Information Mining IIM) pour Mieux comprendre et exploiter des données en provenance du high resolution Synthetic Aperture Radar (SAR) les avancements dans ce champ de recherche contribuent à l'élaboration d'outils d'exploration interactive et l'extraction du contenu de l'image Dans ce contexte , analyser et évaluer les modèles d'image et méthodes d'extraction d'information adéquats selon les conjectures de l'utilisateur, constituent des problèmes difficiles. Notre travail contribue avec des solutions pour la modélisation de SAR de haute résolution et pour l'estimation du contenu en utilisant une approche d'évaluation pilotés par les données (data-driven), et avec la conception de scénarios pour l'extraction d'information dans des images en y associant l'utilisateur et ses conjectures, réalisée par une approche d'évaluation axée sur l'utilisateur. Pour représenter les données et pour permettre l'extraction de l'information, nous nous concentrons sur les propriétés des images à haute résolution SAR et comment les modèles stochastiques peuvent représenter et caractériser le contenu de l'image après une étape d'estimation des paramètres. Nous réalisons une évaluation et une validation guidée par les données des méthodes d'extraction automatique d'informations pour des scènes en haute résolution SAR basée sur le modèle Gibbs Random Field (GRF). Plus précisément, des modèles Gauss Markov Random Field (GMRF) et Auto-binomial (ABM) sont mis en place dans les méthodes d'extraction d'information suite aux deux niveaux d'inférence bayésienne: ajustement du modèle et sélection du modèle. Les deux méthodes donnent comme résultat une image sans tache (speckle-free) et ses paramètres de la structure. Afin d'évaluer la qualité de ces méthodes, nous réalisons des tests de détection sur des classes telles que les villes, la végétation et des plans d'eau; en utilisant des paramètres qualitatifs spécifiques pour quantifier la qualité de l'enlèvement de la tâche (speckle) La précision de la modélisation et la caractérisation du contenu de l'image sont déterminées en utilisant des classifications supervisées et non supervisées, et les matrices de confusion. Nous concluons que les deux méthodes améliorent l'image pendant le processus de nettoyage de l'image. Le modèle GMRF est plus approprié pour les scènes naturelles et le modèle ABM pour les structures artificielles (man-made). Toutefois, l'évaluation des méthodes d'extraction d'information ne suffit pas pour une validation complète de systèmes de type IIM, parce que nous devons nous adapter aux conjectures de l'utilisateur en créant des scénarios de validation et en évaluant le degré de satisfaction des utilisateurs ainsi comme l'efficacité du processus de récupération. Nous concevons et générons deux cas d'étude, qui reflètent les besoins des utilisateurs dans la résolution rapide d'applications de cartographie. L'utilisateur final est inclus dans la méthode d'évaluation en créant deux scénarios d'évaluation dans le cadre de la surveillance des catastrophes: détection des déversements de pétrole et d'inondation. Les scénarios sont effectués en utilisant des produits ScanSAR et High Resolution Spotlight TerraSAR-X, respectivement. Les métriques quantitatives comme la précision et le rappel (recall) sont utilisés comme facteurs de qualité Afin d'avoir des mesures sur le degré de satisfaction des utilisateurs, un groupe d'évaluateurs sont invités à classer de façon qualitative les résultats récupérés. Nous concluons que l'efficacité du processus de récupération est supérieure à 80 pour cent et le degré de satisfaction des utilisateurs est bonne pour les deux scénarios.

This thesis addresses the problem of refocusing smeared SAR images caused by ionosphere phase scintillation noise. A SAR data is smeared when the received signal experiences phase irregularities due to platform orbit deviation, target movement, or, in this thesis, ionospheric scintillation noise due to trans-ionosphere propagation is analyzed. A SAR simulator is constructed to generate stripmap and spotlight data that satisfy theoretically predicted performances under ideal conditions. The simulator is incorporated with ionospheric phase scintillation models to analyze the broadening effect on system’s PSF. Degraded simulation spotlight data are used to test the proposed compensation algorithm. This thesis proposes a two-dimensional polynomial phase fitting algorithm to compensate scintillation noise. This work discusses some requirements of the scene in order to carry out the compensation and what is gained and lost in the process. A successful application of the proposed algorithm to TerraSAR-X data is also presented.

This thesis addresses the problem of refocusing smeared SAR images caused by ionosphere phase scintillation noise. A SAR data is smeared when the received signal experiences phase irregularities due to platform orbit deviation, target movement, or, in this thesis, ionospheric scintillation noise due to trans-ionosphere propagation is analyzed. A SAR simulator is constructed to generate stripmap and spotlight data that satisfy theoretically predicted performances under ideal conditions. The simulator is incorporated with ionospheric phase scintillation models to analyze the broadening effect on system’s PSF. Degraded simulation spotlight data are used to test the proposed compensation algorithm. This thesis proposes a two-dimensional polynomial phase fitting algorithm to compensate scintillation noise. This work discusses some requirements of the scene in order to carry out the compensation and what is gained and lost in the process. A successful application of the proposed algorithm to TerraSAR-X data is also presented.

Spaceborne Synthetic Aperture Radar (SAR) is commonly considered a powerful sensor to detect sea ice. Unfortunately, the sea-ice types in SAR images are difficult to be interpreted due to speckle noise. SAR image denoising therefore becomes a critical step of SAR sea-ice image processing and analysis. In this study, a two-phase approach is designed and implemented for SAR sea-ice image segmentation. In the first phase, a Gamma-based bilateral filter is introduced and applied for SAR image denoising in the local domain. It not only perfectly inherits the conventional bilateral filter with the capacity of smoothing SAR sea-ice imagery while preserving edges, but also enhances it based on the homogeneity in local areas and Gamma distribution of speckle noise. The Gamma-based bilateral filter outperforms other widely used filters, such as Frost filter and the conventional bilateral filter. In the second phase, the K-means clustering algorithm, whose initial centroids are optimized, is adopted in order to obtain better segmentation results. The proposed approach is tested using both simulated and real SAR images, compared with several existing algorithms including K-means, K-means based on the Frost filtered images, and K-means based on the conventional bilateral filtered images. The F1 scores of the simulated results demonstrate the effectiveness and robustness of the proposed approach whose overall accuracies maintain higher than 90% as variances of noise range from 0.1 to 0.5. For the real SAR images, the proposed approach outperforms others with average overall accuracy of 95%.

Part I: Various image reconstruction algorithms used for subsurface targets are reviewed. It is shown how some approximate wavefield inversion techniques: Stripmap Synthetic Aperture Radar (SAR), Kirchhoff Migration (KM) and Frequency-Wavenumber (FK) migration are developed from various models for wavefield scattering. The similarities of these techniques are delineated both from a theoretical and practical perspective and it is shown that Stripmap SAR is, computationally, almost identical to FK migration. A plane wave interpretation of both Stripmap SAR and FK migration is used to show why they are so similar. The electromagnetic assumptions made in the image reconstruction algorithms are highlighted. In addition, it is shown that, theoretically, FK and KM are identical. Image reconstruction results for KM, Stripmap SAR and FK are shown for both synthetic and experimental Ground Penetrating Radar (GPR) data. Subjectively the reconstructed images show little difference, but computationally, Stripmap SAR (and therefore, FK migration) are much more efficient. Part II: A preliminary investigation into the use of the Singularity Expansion Method (SEM) for use in identifying landmines is completed using a Finite-Difference Time-Domain code to simulate a simplified GPR system. The Total Least Squares Matrix Pencil Method (TLS-MPM) is used to determine the complex poles from an arbitrary late-time signal. Both dielectric and metallic targets buried in lossless and lossy half-spaces are considered. Complex poles (resonances) of targets change significantly when the objects are buried in an external medium, and perturbation formulae for Perfect Electric Conductor (PEC) and dielectric targets are highlighted and used. These perturbation formulae are developed for homogenous surrounding media, and their utilization for the half-space (layered medium) GPR problem causes inaccuracies in their predictions. The results show that the decay rate (real part) of the complex poles is not suitable for identification in this problem, but that with further research, the resonant frequency (imaginary part) of the complex poles shows promise as an identification feature. / February 2005

This report treats detection in IR-pictures taken from airplanes over the ground. The detection is divided in two parts. First there is a detection with filterkernels with the task to point on targets and objects that look like targets. The second part is a discriminator that demands more calculations and has the task to sort out the false alarms from the discriminator. Both the detector and the discriminator contain thresholds thats been trained from trainingsets of data. The results from the detector was better then expected hence it wasn’t possible to test the diskriminator properly. / Den här rapporten behandlar detektering i IR-bilder som är tagna från flygplan över marken. Detekteringen är i det här arbetet uppdelat i två delar. Först görs en detektering med filterkärnor som har till uppgift att peka ut mål och målliknande objekt. Den andra delen är en diskriminator som är något mer beräkningskrävande och har till uppgift att sortera bort de falskmål som detektorn fått med. Både detektorn och diskriminatorn innehåller trösklar som har tränats fram med träningsdata. Detektorn gav bättre resultat än väntat och diskriminatorn har därför inte kunnat testas ordentligt.

The first part of the thesis consists of a brief introduction to the general principles of target detection and the sensor-systems used. In the following part there is a theoretical description of the algorithms this thesis focuses on. The detection algorithms described in this paper are called Cell Average, Ordered Statistics, 2parameter and Gammadetector. Two different discriminators called Extended Fractal Features and Quadratic Gamma Discriminator are also described. The algorithms are tested on three different types of data, simulated SAR-pictures, authentic SAR-targets and IR-pictures. The last part account for the results, both those achieved with pictures and those results achieved when doing statistical tests, in this case MonteCarlo- simulations and Reciever Operating Characteristics-curves. The results show that the Gamma- detector and the QGD-algorithm perform best on the tests done in this thesis.

Spatial intensity variation introduced by illumination changes is a challenging problem for image segmentation and classification. Many techniques have been proposed which focus on removing this illumination variation by estimating or modelling it. There is limited research on developing an illumination invariant classification technique which does not use any preprocessing. A major focus of this research is on automatically classifying synthetic aperture radar (SAR) images. These are large satellite images which pose many challenges for image classification including the incidence angle effect which is a strong illumination variation across the image. Mapping of full scene satellite images of sea-ice is important for navigational purposes for ships and also for climate research. The images obtained from the RADARSAT-2 satellite are dual band, high quality images. Currently, sea ice chart are produced manually by ice analysts at the Canadian Ice Service. However, this process can be automated to reduce processing time and obtain more detailed pixel-level ice maps. An automated classification algorithm to achieve sea ice and open water separation will greatly help the ice analyst by providing sufficient guidance in the initial stages of creating an ice map. It would also help the analyst to improve the accuracy while finding ice concentrations and remove subjective bias. The existing Iterative Region Growing by Semantics (IRGS) algorithm is not effective for full scene segmentation because of the incidence angle effect. This research proposes a "glocal" (global as well as local) approach to solve this problem. The image is divided in a rectangular grid and each rectangle is segmented using IRGS. This is viewed as an over-segmentation of the original image. Finally, IRGS is used globally to glue together the over-segmented regions. This method yields acceptable results with the denoised images. The proposed technique can also be used for general image classification purposes. Extensive testing was done to investigate the best set of parameters for the proposed approach. Images were simulated with the SAR illumination variation and multiplicative speckle noise. The technique was effective for general classification and attained accurate results for full scene SAR segmentation.