Investigação sobre métodos para redução de dimensionalidade dos dados em imagens hiperespectrais

Os recentes avanços na tecnologia de sensores tem disponibilizado imagens em alta dimensionalidade para fins de sensoriamento Remoto. Análise e interpretação dos dados provenientes desta nova geração de sensores apresenta novas possibilidades e também novos desafios. Neste contexto,um dos maiores desafios consiste na estimação dos parâmetros em um classificador estatístico utilizando-se um número limitado de amostras de treinamento.Neste estudo,propõe-se uma nova metodologia de extração de feições para a redução da dimensionalidadedos dados em imagens hiperespectrais. Essa metodologia proposta é de fácil implementação e também eficiente do ponto de vista computacional.A hipótese básica consiste em assumir que a curva de resposta espectral do pixel, definida no espaço espectral, pelos contadores digitais (CD's) das bandas espectrais disponíveis, pode ser substituída por um número menor de estatísticas, descrevendo as principais característicasda resposta espectral dos pixels. Espera-se que este procedimento possa ser realizado sem uma perda significativa de informação. Os CD's em cada banda espectral são utilizados para o cálculo de um número reduzido de estatísticas que os substituirão no classificador. Propõe-se que toda a curva seja particionada em segmentos, cada segmento sendo então representado pela respectiva média e variância dos CD's. Propõem-se três algoritmos para segmentação da curva de resposta espectral dos pixels. O primeiro utiliza um procedimento muito simples. Utilizam-se segmentos de comprimento constante, isto é, não se faz nenhuma tentativa para ajustar o comprimento de cada segmento às características da curva espectral considerada. Os outros dois implementam um método que permite comprimentos variáveis para cada segmento,onde o comprimentodos segmentos ao longo da curva de resposta espectral é ajustado seqüencialmente.Um inconveniente neste procedimento está ligado ao fato de que uma vez selecionadauma partição, esta não pode ser alterada, tornando os algoritmos sub-ótimos. Realizam-se experimentos com um classificador paramétrico utilizando-se uma imagem do sensor AVIRIS. Obtiveram-se resultados animadores em termos de acurácia da classificação,sugerindo a eficácia dos algoritmos propostos.

Um método projetivo para cálculo de dimensões de caixas em tempo real

Neste trabalho é descrito um método automático para o cálculo das dimensões de caixas, em tempo real, a partir de uma única imagem obtida com projeção perspectiva. Conhecendo a orientação da caixa no espaço tridimensional e sua distância em relação à câmera, as coordenadas 3D de seus vértices podem ser estimadas e suas dimensões calculadas. Na técnica proposta, são utilizados conceitos de geometria projetiva para estimar a orientação espacial da caixa de interesse a partir de sua silhueta. Já a distância da caixa em relação à câmera é estimada por meio da projeção de feixes de laser sobre uma das faces visíveis da caixa. Esta abordagem pode ser aplicada quando duas ou três faces da caixa de interesse são visíveis simultaneamente na imagem, mesmo quando a caixa encontra-se parcialmente oclusa por outros objetos na cena. Entre as contribuições deste trabalho está o desenvolvimento de um eficiente processo de votação para a transformada de Hough, onde os pixels de uma imagem binária são processados em grupos ao invés de individualmente, como ocorre no método convencional. Também é apresentado um modelo estatístico para a remoção de fundo de cena. Nesse modelo, a cor de fundo é representada sob diferentes condições de iluminação por meio da delimitação de uma região no espaço de cores RGB. O modelo proposto não requer parametrização e é próprio para o uso em aplicações que requeiram câmeras móveis. Para a validação das técnicas descritas neste trabalho, foi construído um protótipo de scanner que calcula as dimensões de caixas a partir de imagens em tempo real. Com o auxilio do scanner, foram capturadas imagens e calculadas as dimensões de diversas caixas reais e sintéticas. As caixas sintéticas foram utilizadas em um ambiente controlado para a validação das técnicas propostas Um dos aspectos importantes deste trabalho é a análise da confiabilidade das medidas obtidas por meio da técnica proposta. Com o objetivo de estudar a propagação de erros ao longo do processo de cálculo das medidas, foi aplicado um método analítico baseado na Teoria de Erros. Também são apresentados estudos estatísticos envolvendo medições realizadas com o protótipo. Estes estudos levam em conta a diferença entre as medidas calculadas pelo sistema e as medidas reais das caixas. A análise dos resultados permite concluir que o método proposto é acurado e preciso.

Uma Fundamentação Intervalar Aplicada à Morfologia Matemática

This work present a interval approach to deal with images with that contain uncertainties, as well, as treating these uncertainties through morphologic operations. Had been presented two intervals models. For the first, is introduced an algebraic space with three values, that was constructed based in the tri-valorada logic of Lukasiewiecz. With this algebraic structure, the theory of the interval binary images, that extends the classic binary model with the inclusion of the uncertainty information, was introduced. The same one can be applied to represent certain binary images with uncertainty in pixels, that it was originated, for example, during the process of the acquisition of the image. The lattice structure of these images, allow the definition of the morphologic operators, where the uncertainties are treated locally. The second model, extend the classic model to the images in gray levels, where the functions that represent these images are mapping in a finite set of interval values. The algebraic structure belong the complete lattices class, what also it allow the definition of the elementary operators of the mathematical morphology, dilation and erosion for this images. Thus, it is established a interval theory applied to the mathematical morphology to deal with problems of uncertainties in images

Técnicas visuais de localização e mapeamento simultâneos sem extração de primitivas geométricas da imagem

In Simultaneous Localization and Mapping (SLAM - Simultaneous Localization and Mapping), a robot placed in an unknown location in any environment must be able to create a perspective of this environment (a map) and is situated in the same simultaneously, using only information captured by the robot s sensors and control signals known. Recently, driven by the advance of computing power, work in this area have proposed to use video camera as a sensor and it came so Visual SLAM. This has several approaches and the vast majority of them work basically extracting features of the environment, calculating the necessary correspondence and through these estimate the required parameters. This work presented a monocular visual SLAM system that uses direct image registration to calculate the image reprojection error and optimization methods that minimize this error and thus obtain the parameters for the robot pose and map of the environment directly from the pixels of the images. Thus the steps of extracting and matching features are not needed, enabling our system works well in environments where traditional approaches have difficulty. Moreover, when addressing the problem of SLAM as proposed in this work we avoid a very common problem in traditional approaches, known as error propagation. Worrying about the high computational cost of this approach have been tested several types of optimization methods in order to find a good balance between good estimates and processing time. The results presented in this work show the success of this system in different environments

Análise da imagem digital para estimativa da área foliar em plantas de laranja Pêra

Objetivou-se, com o trabalho, avaliar dois métodos de estimativa da área foliar, em plantas de laranja Pêra, pela análise da imagem digital obtida com scanner e câmera fotográfica digital. Para determinar a área das folhas, um grupo de discos foi colocado sobre um leitor de scanner, sendo que a imagem obtida foi armazenada. Os mesmos grupos de discos foram fixados sobre cartolina branca e fotografados com câmera fotográfica digital. As imagens obtidas da câmera fotográfica e do scanner foram processadas utilizando ferramentas de um editor de imagem que permite a contagem de pixels de determinada cor, no caso verde. Para a comparação dos métodos, os discos foram submetidos a integrador óptico de área foliar modelo LICOR-3100, utilizando os mesmos agrupamentos. Foram coletadas 20 folhas (cinco em cada quadrante da planta) por parcela de um experimento para comparação de fertilizantes comerciais e doses de zinco, aplicados via foliar, em plantas de sete anos de idade. O experimento foi composto de sete tratamentos e quatro repetições, num total de 28 parcelas. Os dois métodos apresentaram alta correlação com a área estimada pelo integrador óptico de área, considerado como método de referência. O método da análise da imagem obtida com câmera fotográfica, na resolução de 5.0 megapixel, foi mais precisa quando comparada à área estimada pelo integrador óptico de área.

Desenvolvimento de um Sistema de Visão Global para uma Frota de Mini-Robôs Móveis

Navigation based on visual feedback for robots, working in a closed environment, can be obtained settling a camera in each robot (local vision system). However, this solution requests a camera and capacity of local processing for each robot. When possible, a global vision system is a cheapest solution for this problem. In this case, one or a little amount of cameras, covering all the workspace, can be shared by the entire team of robots, saving the cost of a great amount of cameras and the associated processing hardware needed in a local vision system. This work presents the implementation and experimental results of a global vision system for mobile mini-robots, using robot soccer as test platform. The proposed vision system consists of a camera, a frame grabber and a computer (PC) for image processing. The PC is responsible for the team motion control, based on the visual feedback, sending commands to the robots through a radio link. In order for the system to be able to unequivocally recognize each robot, each one has a label on its top, consisting of two colored circles. Image processing algorithms were developed for the eficient computation, in real time, of all objects position (robot and ball) and orientation (robot). A great problem found was to label the color, in real time, of each colored point of the image, in time-varying illumination conditions. To overcome this problem, an automatic camera calibration, based on clustering K-means algorithm, was implemented. This method guarantees that similar pixels will be clustered around a unique color class. The obtained experimental results shown that the position and orientation of each robot can be obtained with a precision of few millimeters. The updating of the position and orientation was attained in real time, analyzing 30 frames per second

Compressão Seletiva de Imagens Coloridas com Detecção Automática de Regiões de Interesse

There has been an increasing tendency on the use of selective image compression, since several applications make use of digital images and the loss of information in certain regions is not allowed in some cases. However, there are applications in which these images are captured and stored automatically making it impossible to the user to select the regions of interest to be compressed in a lossless manner. A possible solution for this matter would be the automatic selection of these regions, a very difficult problem to solve in general cases. Nevertheless, it is possible to use intelligent techniques to detect these regions in specific cases. This work proposes a selective color image compression method in which regions of interest, previously chosen, are compressed in a lossless manner. This method uses the wavelet transform to decorrelate the pixels of the image, competitive neural network to make a vectorial quantization, mathematical morphology, and Huffman adaptive coding. There are two options for automatic detection in addition to the manual one: a method of texture segmentation, in which the highest frequency texture is selected to be the region of interest, and a new face detection method where the region of the face will be lossless compressed. The results show that both can be successfully used with the compression method, giving the map of the region of interest as an input

Caracterização espectral da água do reservatório de Itupararanga, SP, a partir de imagens hiperespectrais hyperion e análise derivativa

The current accessibility to hyperspectral images of Hyperion/EO1 orbital sensor has brought new perspectives for studies of aquatic environments for allowing the remote estimative of several optically active constituents (OACs) in water body. The changes in the composition and concentration of OACs cause different patterns of absorption and scattering of electromagnetic radiation, likely to be detected using hyperspectral data. Therefore, an investigation was conducted taking into account the spectral characterization of water of a reservoir intended for public supply (Itupararanga Reservoir), from Hyperion/EO1 images and derivative analysis technique applied to spectral curves generated. Simultaneously to the acquisition of a Hyperion/EO1 image, a field campaign was carried out to collect limnological data in situ in georeferenced points. After radiometric correction of the image, reflectance curves of pixels were extracted for each station and the curves obtained were subjected to the technique of derivative analysis, which revealed features of absorption and scattering mainly associated to the presence of algal pigments. The results obtained show the presence of phytoplankton and algal activity, matching the field observation.

Extração de feições retas com precisão subpixel: estudo comparativo entre três métodos

In the fields of Machine Vision and Photogrammetry, extracted straight lines from digital images can be used either as vector elements of a digital representation or as control entities that allow the determination of the camera interior and exterior orientation parameters. Applications related with image orientation require feature extraction with subpixel precision, to guarantee the reliability of the estimated parameters. This paper presents three approaches for straight line extraction with subpixel precision. The first approach considers the subpixel refinement based on the weighted average of subpixel positions calculated on the direction perpendicular to the segmented straight line. In the second approach, a parabolic function is adjusted to the grey level profile of neighboring pixels in a perpendicular direction to the segmented line, followed by an interpolation of this model to estimate subpixel coordinates of the line center. In the third approach, the subpixel refinement is performed with a parabolic surface adjustment to the grey level values of neighboring pixels around the segmented line. The intersection of this surface with a normal plane to the line direction generates a parabolic equation that allows estimating the subpixel coordinates of the point in the straight line, assuming that this is the critical point of this function. Three experiments with real images were made and the approach based on parabolic surface adjustment has presented better results.

Extração automatica de segmentos lineares de rodovias em imagens de resoluções variadas

In this paper a methodology for automatic extraction of road segments from images with different resolutions (low, middle and high resolution) is presented. It is based on a generalized concept of lines in digital images, by which lines can be described by the centerlines of two parallel edges. In the specific case of low resolution images, where roads are manifested as entities of 1 or 2 pixels wide, the proposed methodology combines an automatic image enhancement operation with the following strategies: automatic selection of the hysteresis thresholds and the Gaussian scale factor; line length thresholding; and polygonization. In medium and high resolution images roads manifest as narrow and elongated ribbons and, consequently, the extraction goal becomes the road centerlines. In this case, it is not necessary to apply the previous enhancement step used to enhance roads in low resolution images. The results obtained in the experimental evaluation satisfied all criteria established for the efficient extraction of road segments from different resolution images, providing satisfactory results in a completely automatic way.

Encoding mechanisms based on fast oscillations in the retina of the cat and their dependencies on anesthesia

Processing in the visual system starts in the retina. Its complex network of cells with different properties enables for parallel encoding and transmission of visual information to the lateral geniculate nucleus (LGN) and to the cortex. In the retina, it has been shown that responses are often accompanied by fast synchronous oscillations (30 - 90 Hz) in a stimulus-dependent manner. Studies in the frog, rabbit, cat and monkey, have shown strong oscillatory responses to large stimuli which probably encode global stimulus properties, such as size and continuity (Neuenschwander and Singer, 1996; Ishikane et al., 2005). Moreover, simultaneous recordings from different levels in the visual system have demonstrated that the oscillatory patterning of retinal ganglion cell responses are transmitted to the cortex via the LGN (Castelo-Branco et al., 1998). Overall these results suggest that feedforward synchronous oscillations contribute to visual encoding. In the present study on the LGN of the anesthetized cat, we further investigate the role of retinal oscillations in visual processing by applying complex stimuli, such as natural visual scenes, light spots of varying size and contrast, and flickering checkerboards. This is a necessary step for understanding encoding mechanisms in more naturalistic conditions, as currently most data on retinal oscillations have been limited to simple, flashed and stationary stimuli. Correlation analysis of spiking responses confirmed previous results showing that oscillatory responses in the retina (observed here from the LGN responses) largely depend on the size and stationarity of the stimulus. For natural scenes (gray-level and binary movies) oscillations appeared only for brief moments probably when receptive fields were dominated by large continuous, flat-contrast surfaces. Moreover, oscillatory responses to a circle stimulus could be broken with an annular mask indicating that synchronization arises from relatively local interactions among populations of activated cells in the retina. A surprising finding in this study was that retinal oscillations are highly dependent on halothane anesthesia levels. In the absence of halothane, oscillatory activity vanished independent of the characteristics of the stimuli. The same results were obtained for isoflurane, which has similar pharmacological properties. These new and unexpected findings question whether feedfoward oscillations in the early visual system are simply due to an imbalance between excitation and inhibition in the retinal networks generated by the halogenated anesthetics. Further studies in awake behaving animals are necessary to extend these conclusions

Padrão de distribuição e caracterização morfológica de fibras serotonérgicas nos núcleos da linha Média/ intralaminares do tálamo do mocó (Kerodon rupestris)

The midline/intralaminar nuclei form a remarkable group of nuclei of the medial and dorsal thalamus. The midline nuclei, in rats, comprises the paratenial nuclei (PT), paraventricular (PV), intermediodorsal (IMD), reuniens (Re) and rhomboid (Rh). The intralaminar nuclei comprises the central medial (CM), paracentral (PC), central lateral (CL) and parafascicular (PF). Such nuclei have dense serotonergic innervation originating from the brainstem, especially from the so-called ascending activation system. These nuclei, in turn, send projections to various cortical and subcortical areas, specifically to limbic areas, which suggests the important role of this neurotransmitter in the limbic circuitry. The aim of this study was to characterize the distribution pattern and morphology of serotonin fibers in the nuclei of the midline and intralaminar thalamic of rocky cavy (Kerodon rupestris), a tipical rodent from brazilizan northeast. To reach this aim we used four rock cavies adults. Following the transcardially perfusion with paraformaldehyde and brain microtomy steps was performed immunohistochemistry for serotonin (5-HT), Nissl technique and subsequent achievement and image analysis to characterize the cytoarchitecture of these nuclei and the serotonergic fibers visualized. An analysis was made of Relative Optical Density (ROD) to semi-quantify the concentration of serotonin fibers in the areas of interest. Thus, we observed a cytoarchitectonic arrangement of these nuclei similar to that found in rats. In case of fibers distribution, those immunoreactive to 5-HT were presented in a higher concentration according as ROD in the midline nuclei relative to intralaminar; Re being the core which has a higher pixel value followed by the PV , Rh, IMD and PT. In intralaminar CL showed higher pixels, followed by nuclei CM, PC and PF. The serotonergic fibers were classified as number of varicosities and axon diameter, therefore find three types of fibers distributed through this nuclear complex: fibers rugous, granular and semi-granular. In PV fibers predominated rugous; in PT fibers predominated granular; IMD, CL and PF fibers were represented by semi-granular and Re, Rh, PC and CM fibers showed granular and semi-granular. Morphological characterization of serotonergic fibers and differences in density between the nuclei may suggest different patterns of synaptic organization of this neurotransmitter beyond confirming his large repertoire functional

Uma investigação de algoritmos exatos e metaheurísticos aplicados ao nonograma

Nonogram is a logical puzzle whose associated decision problem is NP-complete. It has applications in pattern recognition problems and data compression, among others. The puzzle consists in determining an assignment of colors to pixels distributed in a N  M matrix that satisfies line and column constraints. A Nonogram is encoded by a vector whose elements specify the number of pixels in each row and column of a figure without specifying their coordinates. This work presents exact and heuristic approaches to solve Nonograms. The depth first search was one of the chosen exact approaches because it is a typical example of brute search algorithm that is easy to implement. Another implemented exact approach was based on the Las Vegas algorithm, so that we intend to investigate whether the randomness introduce by the Las Vegas-based algorithm would be an advantage over the depth first search. The Nonogram is also transformed into a Constraint Satisfaction Problem. Three heuristics approaches are proposed: a Tabu Search and two memetic algorithms. A new function to calculate the objective function is proposed. The approaches are applied on 234 instances, the size of the instances ranging from 5 x 5 to 100 x 100 size, and including logical and random Nonograms

Enteric coated magnetic HPMC capsules evaluated in human gastrointestinal tract by AC biosusceptometry

Purpose. To employ the AC Biosusceptometry (ACB) technique to evaluate in vitro and in vivo characteristics of enteric coated magnetic hydroxypropyl methylcellulose (HPMC) capsules and to image the disintegration process.Materials and Methods. HPMC capsules filled with ferrite (MnFe2O4) and coated with Eudragit (R) were evaluated using USP XXII method and administered to fasted volunteers. Single and multisensor ACB systems were used to characterize the gastrointestinal (GI) motility and to determine gastric residence time (GRT), small intestinal transit time (SITT) and orocaecal transit time (OCTT). Mean disintegration time (t (50)) was quantified from 50% increase of pixels in the imaging area.Results. In vitro and in vivo performance of the magnetic HPMC capsules as well as the disintegration process were monitored using ACB systems. The mean disintegration time (t (50)) calculated for in vitro was 25 +/- 5 min and for in vivo was 13 +/- 5 min. In vivo also were determined mean values for GRT (55 +/- 19 min), SITT (185 +/- 82 min) and OCTT (240 +/- 88 min).Conclusions. AC Biosusceptometry is a non-invasive technique originally proposed to monitoring pharmaceutical dosage forms orally administered and to image the disintegration process.

Úlceras crônicas dos membros inferiores: avaliação pela fotografia digital

OBJETIVO: Avaliar o desempenho da análise de imagem digital na estimativa da área acometida pelas úlceras crônicas dos membros inferiores. MÉTODOS: Estudo prospectivo em que foram mensuradas úlceras empregando o método planimétrico clássico, utilizando desenho dos seus contornos em filme plástico transparente, medida sua área posteriormente por folha milimetrada. Esses valores foram utilizados como padrão para a comparação com a estimativa de área pelas fotografias digitais padronizadas das úlceras e dos desenhos das mesmas em filme plástico. Para criar um referencial de conversão dos pixels em milímetros, foi empregado um adesivo com tamanho conhecido, adjacente à úlcera. RESULTADOS: foram avaliadas 42 lesões em 20 pacientes portadores de úlceras crônicas de membros inferiores. As áreas das úlceras variaram de 0,24 a 101,65cm². Observou-se forte correlação entre as medidas planimétricas e as fotos das úlceras (R²=0,86 p<0,01), porém a correlação das medidas planimétricas com as fotos digitais dos desenhos das úlceras foi ainda maior (R²=0,99 p<0,01). CONCLUSÃO: A fotografia digital padronizada revelou-se método rápido, preciso e não-invasivo capaz de estimar a área afetada por úlceras. A avaliação das medidas fotográficas dos contornos das úlceras deve ser preferida em relação à análise de sua fotografia direta.