Análise de séries temporais de coordenadas estimadas com GPS: uma proposta metodológica para eliminação de efeitos sazonais

Coordenação de Aperfeiçoamento de Pessoal de Nível Superior (CAPES)

Caminhamento fotogramétrico utilizando o fluxo óptico filtrado

Coordenação de Aperfeiçoamento de Pessoal de Nível Superior (CAPES)

Two-Stage Orthogonal Least Squares Methods for Neural Network Construction

A number of neural networks can be formulated as the linear-in-the-parameters models. Training such networks can be transformed to a model selection problem where a compact model is selected from all the candidates using subset selection algorithms. Forward selection methods are popular fast subset selection approaches. However, they may only produce suboptimal models and can be trapped into a local minimum. More recently, a two-stage fast recursive algorithm (TSFRA) combining forward selection and backward model refinement has been proposed to improve the compactness and generalization performance of the model. This paper proposes unified two-stage orthogonal least squares methods instead of the fast recursive-based methods. In contrast to the TSFRA, this paper derives a new simplified relationship between the forward and the backward stages to avoid repetitive computations using the inherent orthogonal properties of the least squares methods. Furthermore, a new term exchanging scheme for backward model refinement is introduced to reduce computational demand. Finally, given the error reduction ratio criterion, effective and efficient forward and backward subset selection procedures are proposed. Extensive examples are presented to demonstrate the improved model compactness constructed by the proposed technique in comparison with some popular methods.

Nonlinear model structure design and construction using orthogonal least squares and D-optimality design

A very efficient learning algorithm for model subset selection is introduced based on a new composite cost function that simultaneously optimizes the model approximation ability and model robustness and adequacy. The derived model parameters are estimated via forward orthogonal least squares, but the model subset selection cost function includes a D-optimality design criterion that maximizes the determinant of the design matrix of the subset to ensure the model robustness, adequacy, and parsimony of the final model. The proposed approach is based on the forward orthogonal least square (OLS) algorithm, such that new D-optimality-based cost function is constructed based on the orthogonalization process to gain computational advantages and hence to maintain the inherent advantage of computational efficiency associated with the conventional forward OLS approach. Illustrative examples are included to demonstrate the effectiveness of the new approach.

Análise de séries temporais de coordenadas estimadas com GPS: uma proposta metodológica para detecção, remoção e recuperação de efeitos sazonais

GPS active networks are more and more used in geodetic surveying and scientific experiments, as water vapor monitoring in the atmosphere and lithosphere plate movement. Among the methods of GPS positioning, Precise Point Positioning (PPP) has provided very good results. A characteristic of PPP is related to the modeling and/or estimation of the errors involved in this method. The accuracy obtained for the coordinates can reach few millimeters. Seasonal effects can affect such accuracy if they are not consistent treated during the data processing. Coordinates time series analyses have been realized using Fourier or Harmonics spectral analyses, wavelets, least squares estimation among others. An approach is presented in this paper aiming to investigate the seasonal effects included in the stations coordinates time series. Experiments were carried out using data from stations Manaus (NAUS) and Fortaleza (BRFT) which belong to the Brazilian Continuous GPS Network (RBMC). The coordinates of these stations were estimated daily using PPP and were analyzed through wavelets for identification of the periods of the seasonal effects (annual and semi-annual) in each time series. These effects were removed by means of a filtering process applied in the series via the least squares adjustment (LSQ) of a periodic function. The results showed that the combination of these two mathematical tools, wavelets and LSQ, is an interesting and efficient technique for removal of seasonal effects in time series.

Análise de séries temporais de coordenadas estimadas com gps: Uma proposta metodológica para detecção, remoção e recuperação de efeitos sazonais

GPS active networks are more and more used in geodetic surveying and scientific experiments, as water vapor monitoring in the atmosphere and lithosphere plate movement. Among the methods of GPS positioning, Precise Point Positioning (PPP) has provided very good results. A characteristic of PPP is related to the modeling and / or estimation of the errors involved in this method. The accuracy obtained for the coordinates can reach few millimeters. Seasonal effects can affect such accuracy if they are not consistent treated during the data processing. Coordinates time series analyses have been realized using Fourier or Harmonics spectral analyses, wavelets, least squares estimation among others. An approach is presented in this paper aiming to investigate the seasonal effects included in the stations coordinates time series. Experiments were carried out using data from stations Manaus (NAUS) and Fortaleza (BRFT) which belong to the Brazilian Continuous GPS Network (RBMC). The coordinates of these stations were estimated daily using PPP and were analyzed through wavelets for identification of the periods of the seasonal effects (annual and semi-annual) in each time series. These effects were removed by means of a filtering process applied in the series via the least squares adjustment (LSQ) of a periodic function. The results showed that the combination of these two mathematical tools, wavelets and LSQ, is an interesting and efficient technique for removal of seasonal effects in time series.

The influence of subpixel measurement on digital camera calibration

The aim of this work is to evaluate the influence of point measurements in images, with subpixel accuracy, and its contribution in the calibration of digital cameras. Also, the effect of subpixel measurements in 3D coordinates of check points in the object space will be evaluated. With this purpose, an algorithm that allows subpixel accuracy was implemented for semi-automatic determination of points of interest, based on Fõrstner operator. Experiments were accomplished with a block of images acquired with the multispectral camera DuncanTech MS3100-CIR. The influence of subpixel measurements in the adjustment by Least Square Method (LSM) was evaluated by the comparison of estimated standard deviation of parameters in both situations, with manual measurement (pixel accuracy) and with subpixel estimation. Additionally, the influence of subpixel measurements in the 3D reconstruction was also analyzed. Based on the obtained results, i.e., on the quantification of the standard deviation reduction in the Inner Orientation Parameters (IOP) and also in the relative error of the 3D reconstruction, it was shown that measurements with subpixel accuracy are relevant for some tasks in Photogrammetry, mainly for those in which the metric quality is of great relevance, as Camera Calibration.

Estudo de métodos para classificação e localização precisa de padrões usando um sistema de luz estruturada

Coordenação de Aperfeiçoamento de Pessoal de Nível Superior (CAPES)

Entropia conjunta de espaço e reqüência espacial estimada através da discriminação de estímulos espaciais com luminância e cromaticidade moduladas por funções de Gábor: implicações para o processamento paralelo de informação no sistema visual humano

O objetivo deste estudo foi estimar a entropia conjunta do sistema visual humano no domínio do espaço e no domínio das freqüências espaciais através de funções psicométricas. Estas foram obtidas com testes de discriminação de estímulos com luminância ou cromaticidade moduladas por funções de Gábor. A essência do método consistiu em avaliar a entropia no domínio do espaço, testando-se a capacidade do sujeito em discriminar estímulos que diferiam apenas em extensão espacial, e avaliar a entropia no domínio das freqüências espaciais, testando-se a capacidade do sujeito em discriminar estímulos que diferiam apenas em freqüência espacial. A entropia conjunta foi calculada, então, a partir desses dois valores individuais de entropia. Três condições visuais foram estudadas: acromática, cromática sem correção fina para eqüiluminância e cromática com correção para eqüiluminância através de fotometria com flicker heterocromático. Quatro sujeitos foram testados nas três condições, dois sujeitos adicionais foram testados na condição cromática sem eqüiluminância fina e um sétimo sujeito também fez o teste acromático. Todos os sujeitos foram examinados por oftalmologista e considerados normais do ponto de vista oftálmico, não apresentando relato, sintomas ou sinais de disfunções visuais ou de moléstias potencialmente capazes de afetar o sistema visual. Eles tinham acuidade visual normal ou corrigida de no mínimo 20/30. O trabalho foi aprovado pela Comissão de Ética em Pesquisa do Núcleo de Medicina Tropical da UFPA e obedeceu às recomendações da Declaração de Helsinki. As funções de Gábor usadas para modulação de luminância ou cromaticidade compreenderam redes senoidais unidimensionais horizontais, moduladas na direção vertical, dentro de envelopes gaussianos bidimensionais cuja extensão espacial era medida pelo desvio padrão da gaussiana. Os estímulos foram gerados usando-se uma rotina escrita em Pascal num ambiente Delphi 7 Enterprise. Foi utilizado um microcomputador Dell Precision 390 Workstation e um gerador de estímulos CRS VSG ViSaGe para exibir os estímulos num CRT de 20”, 800 x 600 pixels, 120 Hz, padrão RGB, Mitsubishi Diamond Pro 2070SB. Nos experimentos acromáticos, os estímulos foram gerados pela modulação de luminância de uma cor branca correspondente à cromaticidade CIE1931 (x = 0,270; y = 0,280) ou CIE1976 (u’ = 0,186; v’= 0,433) e tinha luminância média de 44,5 cd/m2. Nos experimentos cromáticos, a luminância média foi mantida em 15 cd/m2 e foram usadas duas series de estímulos verde-vermelhos. Os estímulos de uma série foram formados por duas cromaticidades definidas no eixo M-L do Espaço de Cores DKL (CIE1976: verde, u’=0,131, v’=0,380; vermelho, u’=0,216, v’=0,371). Os estímulos da outra série foram formados por duas cromaticidades definidas ao longo de um eixo horizontal verde-vermelho definido no Espaço de Cores CIE1976 (verde, u’=0,150, v’=0,480; vermelho, u’=0,255, v’=0,480). Os estímulos de referência eram compostos por redes de três freqüências espaciais diferentes (0,4, 2 e 10 ciclos por grau) e envelope gaussiano com desvio padrão de 1 grau. Os estímulos de testes eram compostos por uma entre 19 freqüências espaciais diferentes em torno da freqüência espacial de referência e um entre 21 envelopes gaussianos diferentes com desvio padrão em torno de 1 grau. Na condição acromática, foram estudados quatro níveis de contraste de Michelson: 2%, 5%, 10% e 100%. Nas duas condições cromáticas foi usado o nível mais alto de contraste agregado de cones permitidos pelo gamut do monitor, 17%. O experimento consistiu numa escolha forçada de dois intervalos, cujo procedimento de testagem compreendeu a seguinte seqüência: i) apresentação de um estímulo de referência por 1 s; ii) substituição do estímulo de referência por um fundo eqüiluminante de mesma cromaticidade por 1 s; iii) apresentação do estímulo de teste também por 1 s, diferindo em relação ao estímulo de referência seja em freqüência espacial, seja em extensão espacial, com um estímulo sonoro sinalizando ao sujeito que era necessário responder se o estímulo de teste era igual ou diferente do estímulo de referência; iv) substituição do estímulo de teste pelo fundo. A extensão espacial ou a freqüência espacial do estímulo de teste foi mudada aleatoriamente de tentativa para tentativa usando o método dos estímulos constantes. Numa série de 300 tentativas, a freqüencia espacial foi variada, noutra série também de 300 tentativas, a extensão espacial foi variada, sendo que cada estímulo de teste em cada série foi apresentado pelo menos 10 vezes. A resposta do indivíduo em cada tentativa era guardada como correta ou errada para posterior construção das curvas psicométricas. Os pontos experimentais das funções psicométricas para espaço e freqüência espacial em cada nível de contraste, correspondentes aos percentuais de acertos, foram ajustados com funções gaussianas usando-se o método dos mínimos quadrados. Para cada nível de contraste, as entropias para espaço e freqüência espacial foram estimadas pelos desvios padrões dessas funções gaussianas e a entropia conjunta foi obtida multiplicando-se a raiz quadrada da entropia para espaço pela entropia para freqüência espacial. Os valores de entropia conjunta foram comparados com o mínimo teórico para sistemas lineares, 1/4π ou 0,0796. Para freqüências espaciais baixas e intermediárias, a entropia conjunta atingiu níveis abaixo do mínimo teórico em contrastes altos, sugerindo interações não lineares entre dois ou mais mecanismos visuais. Este fenômeno occorreu em todas as condições (acromática, cromática e cromática eqüiluminante) e foi mais acentuado para a frequência espacial de 0,4 ciclos / grau. Uma possível explicação para este fenômeno é a interação não linear entre as vias visuais retino-genículo-estriadas, tais como as vias K, M e P, na área visual primária ou em níveis mais altos de processamento neural.

Comparison of partial least squares regression and principal component regression for pelvic shape prediction.

This paper studied two different regression techniques for pelvic shape prediction, i.e., the partial least square regression (PLSR) and the principal component regression (PCR). Three different predictors such as surface landmarks, morphological parameters, or surface models of neighboring structures were used in a cross-validation study to predict the pelvic shape. Results obtained from applying these two different regression techniques were compared to the population mean model. In almost all the prediction experiments, both regression techniques unanimously generated better results than the population mean model, while the difference on prediction accuracy between these two regression methods is not statistically significant (α=0.01).

A proposal for the metrological characterization of circular features with digital optical machines

This paper aims to analyze the different adjustment methods commonly used to characterize indirect metrology circular features: least square circle, minimum zone circle, maximum inscribed circle and minimum circumscribed circle. The analysis was performed from images obtained by digital optical machines. The calculation algorithms, self-developed, have been implemented in Matlab® and take into consideration as study variables: the amplitude of angular sector of the circular feature, its nominal radio and the magnification used by the optical machine. Under different conditions, it was determined the radius and circularity error of different circular standards. The comparison of the results, obtained by the different methods of adjustments used, with certified values for the standards, has allowed us to determine the accuracy of each method and its scope.

Pricing-to-market for UK export sector

This thesis investigates the pricing-to-market (PTM) behaviour of the UK export sector. Unlike previous studies, this study econometrically tests for seasonal unit roots in the export prices prior to estimating PTM behaviour. Prior studies have seasonally adjusted the data automatically. This study’s results show that monthly export prices contain very little seasonal unit roots implying that there is a loss of information in the data generating process of the series when estimating PTM using seasonally-adjusted data. Prior studies have also ignored the econometric properties of the data despite the existence of ARCH effects in such data. The standard approach has been to estimate PTM models using Ordinary Least Square (OLS). For this reason, both EGARCH and GJR-EGARCH (hereafter GJR) estimation methods are used to estimate both a standard and an Error Correction model (ECM) of PTM. The results indicate that PTM behaviour varies across UK sectors. The variables used in the PTM models are co-integrated and an ECM is a valid representation of pricing behaviour. The study also finds that the price adjustment is slower when the analysis is performed on real prices, i.e., data that are adjusted for inflation. There is strong evidence of auto-regressive condition heteroscedasticity (ARCH) effects – meaning that the PTM parameter estimates of prior studies have been ineffectively estimated. Surprisingly, there is very little evidence of asymmetry. This suggests that exporters appear to PTM at a relatively constant rate. This finding might also explain the failure of prior studies to find evidence of asymmetric exposure in foreign exchange (FX) rates. This study also provides a cross sectional analysis to explain the implications of the observed PTM of producers’ marginal cost, market share and product differentiation. The cross-sectional regressions are estimated using OLS, Generalised Method of Moment (GMM) and Logit estimations. Overall, the results suggest that market share affects PTM positively.Exporters with smaller market share are more likely to operate PTM. Alternatively, product differentiation is negatively associated with PTM. So industries with highly differentiated products are less likely to adjust their prices. However, marginal costs seem not to be significantly associated with PTM. Exporters perform PTM to limit the FX rate effect pass-through to their foreign customers, but they also avoided exploiting PTM to the full, since to do so can substantially reduce their profits.

Decomposition of aggregated load : finding induction motor fraction in real load

The main contribution of this paper is decomposition/separation of the compositie induction motors load from measurement at a system bus. In power system transmission buses load is represented by static and dynamic loads. The induction motor is considered as the main dynamic loads and in the practice for major transmission buses there will be many and various induction motors contributing. Particularly at an industrial bus most of the load is dynamic types. Rather than traing to extract models of many machines this paper seeks to identify three groups of induction motors to represent the dynamic loads. Three groups of induction motors used to characterize the load. These are the small groups (4kw to 11kw), the medium groups (15kw to 180kw) and the large groups (above 630kw). At first these groups with different percentage contribution of each group is composite. After that from the composite models, each motor percentage contribution is decomposed by using the least square algorithms. In power system commercial and the residential buses static loads percentage is higher than the dynamic loads percentage. To apply this theory to other types of buses such as residential and commerical it is good practice to represent the total load as a combination of composite motor loads, constant impedence loads and constant power loads. To validate the theory, the 24hrs of Sydney West data is decomposed according to the three groups of motor models.