The purpose of this project was to test and implement recent research of polarization and scatter properties that suggest using a cross polarization imaging system to reduce glare artifacts. In particular, the use of this research is to improve the machine vision of apple quality detection in the food industry. The automated measurement system was implemented by acquiring pictures at different angles and different polarization states of apples. The opto-mechanics, system integration, synchronization and data collection are controlled with LabVIEW.

Humans recognize optical reflectance properties of surfaces such as metal, plastic, or paper from a single image without knowledge of illumination. We develop a machine vision system to perform similar recognition tasks automatically. Reflectance estimation under unknown, arbitrary illumination proves highly underconstrained due to the variety of potential illumination distributions and surface reflectance properties. We have found that the spatial structure of real-world illumination possesses some of the statistical regularities observed in the natural image statistics literature. A human or computer vision system may be able to exploit this prior information to determine the most likely surface reflectance given an observed image. We develop an algorithm for reflectance classification under unknown real-world illumination, which learns relationships between surface reflectance and certain features (statistics) computed from a single observed image. We also develop an automatic feature selection method.

Under normal viewing conditions, humans find it easy to distinguish between objects made out of different materials such as plastic, metal, or paper. Untextured materials such as these have different surface reflectance properties, including lightness and gloss. With single isolated images and unknown illumination conditions, the task of estimating surface reflectance is highly underconstrained, because many combinations of reflection and illumination are consistent with a given image. In order to work out how humans estimate surface reflectance properties, we asked subjects to match the appearance of isolated spheres taken out of their original contexts. We found that subjects were able to perform the task accurately and reliably without contextual information to specify the illumination. The spheres were rendered under a variety of artificial illuminations, such as a single point light source, and a number of photographically-captured real-world illuminations from both indoor and outdoor scenes. Subjects performed more accurately for stimuli viewed under real-world patterns of illumination than under artificial illuminations, suggesting that subjects use stored assumptions about the regularities of real-world illuminations to solve the ill-posed problem.

The bidirectional reflectance distribution function has a four dimensional parameter space and such high dimensionality makes it impractical to use it directly in hardware rendering. When a BRDF has no analytical representation, common solutions to overcome this problem include expressing it as a sum of basis functions or factorizing it into several functions of smaller dimensions. This thesis describes factorization extensions that significantly improve factor computation speed and eliminate drawbacks of previous techniques that overemphasize low sample values. The improved algorithm is used to calculate factorizations and material maps from colored images. The technique presented in this thesis allows interactive definition of arbitrary materials, and although this method is based on physical parameters, it can be also used for achieving a variety of non-photorealistic effects.

In this thesis, a modified BRDF model for wide-angle scattering is presented. The proposed model is developed from empirical observations of several BRDF models. The model is an extention of the classical microfacet models. By replacing the two cosines of elevation angles with functions and exponent parameters, our model is able to give a special characteristic which we have not found in any other BRDF models. The characteristic at wide-angle scattering can be, for example, seen on the polyethylene material. In addition, our proposed model can greatly improve relative error from the reference model. The average relative error improvement is about 20 ercent for a cosine weighted error metric,E1 , and 10 percent for a logarithmic error metric, E 2,. Moreover, we also introduce a new optimization approach for the proposed terms. This approach can do optimization so that our proposed model gives at least an equivalent error to the reference model.

The capability to map anthropogenic aerosol quantities and properties over land can provide significant insights for climate and environmental studies on global and regional scales. One of the primary challenges in aerosol information monitoring is separating two signals measured by downward-viewing airborne or spaceborne instruments: the light scattered from the aerosols and light reflected from the Earth's surface. In order to study the aerosols independently, the surface signal needs to be subtracted out from the measurements. Some observational modalities, such as multispectral and multiangle, do not provide enough information to uniquely define the Earth's directional reflectance properties for this task due to the high magnitude and inhomogeneity of albedo for land surface types. Polarization, however, can provide additional information to define surface reflection. To improve upon current measurement capabilities of aerosols over urban areas, Jet Propulsion Laboratory developed the Multiangle SpectroPolarimetric Imager (MSPI) that can accurately measure the Degree of Linear Polarization to 0.5%. In particular, data acquired by the ground-based prototype, GroundMSPI, is used for directional reflectance studies of outdoor surfaces in this dissertation. This work expands upon an existing model, the microfacet model, to characterize the polarized bidirectional reflectance distribution function (pBRDF) of surfaces and validate an assumption, the Spectral Invariance Hypothesis, on the surface pBRDF that is used in aerosol retrieval algorithms. The microfacet model is commonly used to represent the pBRDF of Earth's surface types, such as ocean and land. It represents a roughened surface comprised of randomly oriented facets that specularly reflect incoming light into the upward hemisphere. The analytic form of the pBRDF for this model assumes only a single reflection of light from the microfaceted surface. If the incoming illumination is unpolarized, as it is with natural light from the Sun, the reflected light is linearly polarized perpendicular to the plane that contains the illumination and view directions, the scattering plane. However, previous work has shown that manmade objects, such as asphalt and brick, show a polarization signature that differs from the single reflection microfacet model. Using the polarization ray-tracing (PRT) program POLARIS-M, a numerical calculation for the pBRDF is made for a roughened surface to account for multiple reflections that light can experience between microfacets. Results from this numerical PRT method shows rays that experience two or more reflections with the microfacet surface can be polarized at an orientation that differs from the analytical single reflection microfacet model. This PRT method is compared against GroundMSPI data of manmade surfaces. An assumption made regarding the pBRDF for this microfacet model is verified with GroundMSPI data of urban areas. This is known as the Spectral Invariance Hypothesis and asserts that the magnitude and shape of the polarized bidirectional reflectance factor (pBRF) is the same for all wavelengths. This simplifies the microfacet model by assuming some surface parameters such as the index of refraction are spectrally neutral. GroundMSPI acquires the pBRF for five prominent region types, asphalt, brick, cement, dirt, and grass, for day-long measurements on clear sky conditions. Over the course of each day, changing solar position in the sky provides a large range of scattering angles for this study. The pBRF is measured for the three polarimetric wavelengths of GroundMSPI, 470, 660, and 865nm, and the best fit slope of the spectral correlation is reported. This investigation shows agreement to the Spectral Invariance Hypothesis within 10% for all region types excluding grass. Grass measurements show a large mean deviation of 31.1%. This motivated an angle of linear polarization (AoLP) analysis of cotton crops to isolate single reflection cases, or specular reflections, from multiple scattering cases of light in vegetation. Results from this AoLP method show that specular reflections off the top surface of leaves follow the Spectral Invariance Hypothesis.

Land surface albedo describes the proportion of incident solar radiant flux that is reflected from the Earth's surface and therefore is a crucial parameter in modeling and monitoring attempts to capture the current climate, hydrological, and biogeochemical cycles and predict future scenarios. Due to the temporal variability and spatial heterogeneity of land surface albedo, remote sensing offers the only realistic method of monitoring albedo on a global scale. While the distribution of bright, highly reflective surfaces (clouds, snow, deserts) govern the vast majority of the fluctuation, variations in the intrinsic surface albedo due to natural and human disturbances such as urban development, fire, pests, harvesting, grazing, flooding, and erosion, as well as the natural seasonal rhythm of vegetation phenology, play a significant role as well. The development of times series of global snow-free and cloud-free albedo from remotely sensed observations over the past decade and a half offers a unique opportunity to monitor and assess the impact of these alterations to the Earth's land surface. By utilizing multiple satellite records from the MODerate-resolution Imaging Spectroradiometer (MODIS), the Multi-angle Imaging Spectroradiometer (MISR) and the Visible Infrared Imaging Radiometer Suite (VIIRS) instruments, and developing innovative spectral conversion coefficients and temporal gap-filling strategies, it has been possible to utilize the strengths of the various sensors to improve the spatial and temporal coverage of global land surface albedo retrievals. The availability of these products is particularly important in tropical regions where cloud cover obscures the forest for significant periods. In the Amazon, field ecologists have noted that some areas of the forest ecosystem respond rapidly with foliage growth at the beginning of the dry season, when sunlight can finally penetrate fully to the surface and have suggested this phenomenon can continue until reductions in water availability (particularly in times of drought) impact the growth cycle. While it has been difficult to capture this variability from individual optical satellite sensors, the temporally gap-filled albedo products developed during this research are used in a case study to monitor the Amazon during the dry season and identify the extent of these regions of foliage growth.

In the development of autonomous vehicles, replacing the driver and its perceptive abilities is one of many technical challenges. As a part in solving these challenges, Light Detection And Ranging (LiDAR) is a promising technology. In short, LiDAR works by using lasers to detect objects in its vicinity by detecting the light that reflects on them. With knowledge of the reflective properties of an object, a prediction can be made regarding whether a certain LiDAR unit will be able to detect the object or not. When making this prediction, the common description of reflectance is often insufficient. Instead, a more complete description is given by the Bidirectional Reflectance Distribution Function (BRDF) of a surface, which describes reflection on the surface while taking the incident and reflected direction into consideration. In this thesis, an experimental setup was built with the capabilities of measuring the BRDF while taking incident and reflected polarization into account. Program software was written in Python and integrated with the hardware, providing a user interface for simple control of the setup. The BRDF was measured on a total of 6 samples; 2 reference samples and 4 samples taken from the hood of 4 different cars. Conclusively, the setup provided useful information about the reflective and polarization properties of the samples. These measurements can help in predicting whether or not a surface can be detected by a given LiDAR unit, and can also be helpful when designing new LiDAR units by providing useful information about the surfaces they are required to detect.

L'albédo, fraction de rayonnement réfléchi dans le spectre solaire, est une variable clef du bilan énergétique des surfaces enneigées et englacées. Cette grandeur possède une forte variabilité spatio-temporelle ce qui fait de la télédétection un outil adapté pour son étude. L'albédo dépend à la fois des propriétés physiques du milieu considéré et des caractéristiques du rayonnement incident. Les différentes grandeurs liées à l'albédo sont fonction des domaines angulaires et spectraux des radiations considérées. Les mesures de répartition angulaire du rayonnement réfléchi par la neige ont montré que l'hypothèse lambertienne pouvait conduire à des erreurs non négligeables lors de la détermination de l'albédo par télédétection. La connaissance des caractéristiques de la répartition angulaire du rayonnement réfléchi par la neige permet de développer une nouvelle méthode de détermination de l'albédo en zones montagneuses. Cette méthode prend en compte les effets liés à la forte variabilité topographique des terrains de montagne, à l'anisotropie du rayonnement réfléchi par la neige et par la glace ainsi que les variations spectrales de l'albédo en fonction des propriétés physiques de la surface. Elle a été appliquée à deux types de données : des photographies terrestres visibles et proche infrarouges (résolution spatiale 10 m) et des images MODIS (résolution spatiale 250 m). L'incertitude sur la valeur de l'albédo ainsi déterminée est évaluée à ±10% grâce aux mesures de terrain effectuées sur le glacier de Saint Sorlin (massif des Grandes Rousses, France). L'étude des cartes d'albédo issues de dix années (2000-2009) d'images MODIS montre qu'il n'y a pas de décroissance marquée de la valeur de l'albédo en zone d'ablation au contraire de ce qui a été prouvé pour le glacier du Morteratsch (Suisse). De plus, il existe une corrélation très élevée entre la valeur minimale de la moyenne de l'albédo sur le glacier, i.e. l'albédo moyen du glacier le jour où la ligne de neige est proche de la ligne d'équilibre, et la valeur du bilan de masse annuel spécifique. L'assimilation des données d'albédo obtenues grâce aux images MODIS et aux photographies terrestres dans le modèle de neige CROCUS permet une bonne estimation du bilan de masse spatialisé du glacier de Saint Sorlin (rmse=0.5 m w.e. pour les cinq années hydrologiques étudiées). Les forçages météorologiques utilisés pour cette étude sont de moyenne échelle. L'analyse succincte de la contribution des différents flux atmosphériques au bilan d'énergie de surface montre qu'en zone d'ablation comme en zone d'accumulation, le bilan radiatif net courtes longueurs d'ondes constitue la source principale d'énergie et que la variabilité de ce flux explique la majeure partie de la variabilité journalière de la somme des flux atmosphériques. Appliquées à d'autres glaciers, ces méthodes permettraient de savoir si les conclusions établies pour notre seul glacier d'étude sont valables pour d'autres glaciers. Elles rendraient également possibles la reconstruction du bilan de masse spatialisé sur 10 ans d'autres glaciers et potentiellement une meilleure quantification des processus physiques mis en jeu dans le bilan de masse de ces glaciers tempérés / Albedo is defined as the ratio of reflected to incident radiation over the solar spectrum and is a key parameter in the surface energy balance of snow and ice. This parameter is highly variable both temporally and spatially; thus remote sensing is an ideally suited approach for the retrieval of albedo data.The albedo value depends on both physical properties of the target and the characteristics of the incident radiation. Furthermore, the physical parameters linked with the albedo concept vary in consideration with spectral and angular ranges. Measurements of bi-directional reflectance over natural snow have shown that the Lambertian hypothesis may lead to significant error when estimating albedo from remote sensing data.Detailed knowledge of the angular distribution of radiation reflected by snow allows for the development of a new method to retrieve albedo values for mountainous, snow/ice covered areas. This method takes into account multiple reflections on mountainous areas, anisotropy of radiation reflected by snow and ice, and albedo spectral variations with surface physical properties. The method is applied to visible and near-infrared terrestrial photographs (spatial resolution 10 m) and MODIS data (spatial resolution 250 m). The accuracy of the method is evaluated at ±10% on the retrieved albedo value using concurrent field measurements at theSaint Sorlin Glacier (Grandes Rousses, France) during the summers of 2008 and 2009. The method is used to retrieve albedo data for this glacier from 2000 to 2009. Results indicate that the albedo of the ablation area of the Saint Sorlin Glacier has not shown any decreasing trend over this decade, in opposition to results presented for the Morteratsch Glacier (Switzerland). In addition, the minimal value over the summer period of the whole glacier averaged albedo is highly correlated to the specific annual mass-balance.Albedo data from MODIS and terrestrial photographs are then assimilated into the snow model CROCUS. This assimilation allows for an estimation of the spatialized mass-balance of the Saint Sorlin Glacier over the five studied hydrological years. Root mean square error is evaluated to 0.5 m w.e. For this study, we have used mid-scale meteorological data from SAFRAN. A brief analysis of the contribution from the atmospheric fluxes to the surface energy balance shows that, for the time period considered in this study, the shortwave radiation budget is the main process determining the surface energy balance. Furthermore, variability in shortwave radiation budget explains the major part of the daily variability in surface energy balance.The methods developed in this work are readily applicable to other temperate glaciers. They allows spatialized mass-balance reconstruction on a decadal scale and lead to improved quantification of the physical processes controlling mass-balance in temperate glaciers

Les surfaces nanostructurées permettent d’obtenir des effets colorés gonio-apparents lorsque les nanostructures présentent des dimensions de l’ordre des longueurs d’onde du spectre visible. ces couleurs sont habituellement modélisées par le biais d’interactions de types «interférences» ou «cristaux photoniques» entre le rayonnement lumineux et une structure modèle. dans cette thèse, l’anodisation d’aluminium est utilisée comme méthode de structuration à l’échelle submicronique. cette technique présente l’avantage d’être mature industriellement et de permettre de structurer de grandes surfaces. des effets colorés sont observés même si les structures obtenues ne sont pas parfaitement ordonnées. le but de cette thèse est de comprendre les phénomènes optiques mis en jeu dans l’obtention de ces effets. ce manuscrit se divise donc en deux parties principales, toutes deux basées sur une étude de la littérature existante. afin d’établir un parallèle entre caractérisation expérimentale et simulation numérique, la première partie présente l’outil de caractérisation optique développé. la seconde est dédiée à l’étude des effets colorés de certaines surfaces d’aluminium anodisé. cette partie propose une compréhension des phénomènes d’interaction de la lumière avec la structure d’aluminium anodisé se basant sur les caractérisations optiques et microstructurales des échantillons, associées à une modélisation de l’interaction entre rayonnement et matière structurée. cette étude montre que les structures réelles présentent un ordre à courte distance. les effets colorés sont simulés par la méthode modale de fourier par le biais de structures modèles avec un certain niveau de désordre. / Nano-structured surfaces allow obtaining of colored gonio-apparent effects when the nano-structures dimensions are of the order of the visible spectrum wavelengths. these colors are usually modeled by means of the interactions type so-called « interferences type » ou « photonic crystals type » between the luminous radiation and a model structure. in this thesis, the anodization of aluminum substrates is used to produce surface structures at the submicron scale. this technique is industrially mature and allows to structure large surfaces. color effects were observed even if obtained structures are not perfectly ordered. the aim of this thesis is the understanding of the optical phenomena involved in the production of such effects. this manuscript is divided into two main parts, both based on the existing literature analysis. in order to draw a parallel between experimental characterization and numerical simulations, one part presents the instrumental development of the optical characterization instrument. the second one is dedicated to the study of color effect of certain anodized aluminum surfaces. this part proposes an understanding of the interaction phenomena between the light and the anodized aluminum structure, based on the optical and microstructural characterization of the samples, associated to a modeling of the interaction between light and structured matter. this study shows that such structures present a short-distance order. the color effects are simulated numerically by fourier modal method by the means of model-structures with certain disorder degree.