<p>In this study the flux attenuation due to sensor displacement has been investigated over sea using an extensive set of data from the "Ocean Horizontal Array Turbulence Study". All previous investigations of the flux attenuation have been performed over land.</p><p>A function developed for correcting fluxes in the homogenous surface layer was compared to measured flux attenuation. This investigation revealed the possibility to find new functions describing the flux attenuation when measurements are carried out over sea. From the measured flux attenuation studied here a change in the form of correction functions was required to improve the estimated flux loss. The most significant difference found in this report compared to the previous landbased study Horst (2006) is for stable conditions, where significantly less flux loss is found over sea. Two new functions describing the attenuation due to sensor displacement over sea have been constructed.</p><p>One of these expressions has a discontinuity at z/L = 0. This is supported by measured flux attenuation. A reasonable interpretation is; however, that this discontinuity is caused by two separate turbulence regimes near neutrality on the stable and unstable side respectively. The discontinuity is thus not believed to be an effect merely of stability. A second correction function which is continuous over all stabilities has therefore also been constructed. These two functions and the correction function from Horst (2006) have been compared to measured flux loss. Based on this comparison the continuous correction function is recommended for correcting scalar fluxes measured over sea. It should be noted, however, that this expression only describes the mean attenuation and has been constructed from measurements at 5 and 5.5 m above mean sea level.</p><p>The theoretical basis used in the development of the function for flux attenuation over land allows for a direct link between a spectral shape and the attenuation expression. This link has been preserved for the new expressions presented in this report. The spectral shape corresponding to the continuous correction function has been compared to measured mean cospectra and also to the cospectra from Horst (2006) corresponding to crosswind displacements.</p><p>At a height of 10 m and a sensor displacement of 0.2 m the mean flux attenuation is about 1.3-4% in the stability interval −1 < z/L < 1.5 when using the new correction functions presented in this report.</p>

In this study the flux attenuation due to sensor displacement has been investigated over sea using an extensive set of data from the "Ocean Horizontal Array Turbulence Study". All previous investigations of the flux attenuation have been performed over land. A function developed for correcting fluxes in the homogenous surface layer was compared to measured flux attenuation. This investigation revealed the possibility to find new functions describing the flux attenuation when measurements are carried out over sea. From the measured flux attenuation studied here a change in the form of correction functions was required to improve the estimated flux loss. The most significant difference found in this report compared to the previous landbased study Horst (2006) is for stable conditions, where significantly less flux loss is found over sea. Two new functions describing the attenuation due to sensor displacement over sea have been constructed. One of these expressions has a discontinuity at z/L = 0. This is supported by measured flux attenuation. A reasonable interpretation is; however, that this discontinuity is caused by two separate turbulence regimes near neutrality on the stable and unstable side respectively. The discontinuity is thus not believed to be an effect merely of stability. A second correction function which is continuous over all stabilities has therefore also been constructed. These two functions and the correction function from Horst (2006) have been compared to measured flux loss. Based on this comparison the continuous correction function is recommended for correcting scalar fluxes measured over sea. It should be noted, however, that this expression only describes the mean attenuation and has been constructed from measurements at 5 and 5.5 m above mean sea level. The theoretical basis used in the development of the function for flux attenuation over land allows for a direct link between a spectral shape and the attenuation expression. This link has been preserved for the new expressions presented in this report. The spectral shape corresponding to the continuous correction function has been compared to measured mean cospectra and also to the cospectra from Horst (2006) corresponding to crosswind displacements. At a height of 10 m and a sensor displacement of 0.2 m the mean flux attenuation is about 1.3-4% in the stability interval −1 &lt; z/L &lt; 1.5 when using the new correction functions presented in this report.

This dissertation introduces a new system for handwritten text recognition based on an improved neural network design. Most of the existing neural networks treat mean square error function as the standard error function. The system as proposed in this dissertation utilizes the mean quartic error function, where the third and fourth derivatives are non-zero. Consequently, many improvements on the training methods were achieved. The training results are carefully assessed before and after the update. To evaluate the performance of a training system, there are three essential factors to be considered, and they are from high to low importance priority: 1) error rate on testing set, 2) processing time needed to recognize a segmented character and 3) the total training time and subsequently the total testing time. It is observed that bounded training methods accelerate the training process, while semi-third order training methods, next-minimal training methods, and preprocessing operations reduce the error rate on the testing set. Empirical observations suggest that two combinations of training methods are needed for different case character recognition. Since character segmentation is required for word and sentence recognition, this dissertation provides also an effective rule-based segmentation method, which is different from the conventional adaptive segmentation methods. Dictionary-based correction is utilized to correct mistakes resulting from the recognition and segmentation phases. The integration of the segmentation methods with the handwritten character recognition algorithm yielded an accuracy of 92% for lower case characters and 97% for upper case characters. In the testing phase, the database consists of 20,000 handwritten characters, with 10,000 for each case. The testing phase on the recognition 10,000 handwritten characters required 8.5 seconds in processing time.

No description available.

Various techniques of optimizing the multiple class cross-entropy error function to train single hidden layer neural network classifiers with softmax output transfer functions are investigated on a real-world multispectral pixel-by-pixel classification problem that is of fundamental importance in remote sensing. These techniques include epoch-based and batch versions of backpropagation of gradient descent, PR-conjugate gradient and BFGS quasi-Newton errors. The method of choice depends upon the nature of the learning task and whether one wants to optimize learning for speed or generalization performance. It was found that, comparatively considered, gradient descent error backpropagation provided the best and most stable out-of-sample performance results across batch and epoch-based modes of operation. If the goal is to maximize learning speed and a sacrifice in generalisation is acceptable, then PR-conjugate gradient error backpropagation tends to be superior. If the training set is very large, stochastic epoch-based versions of local optimizers should be chosen utilizing a larger rather than a smaller epoch size to avoid inacceptable instabilities in the generalization results. (authors' abstract) / Series: Discussion Papers of the Institute for Economic Geography and GIScience

Traverses Adjustment in the surface of the ellipsoid with the objectives to guarantee the solution unicity in the transport of curvilinear geodesic coordinates (latitude and longitude) and in the azimuth transport and to get the estimates of quality. It deduces the coordinate transport and the azimuth transport by mean Legendre s series of the geodesic line. This series is based on the Taylor s series, where the argument is the length of the geodesic line. For the practical applications, it has the necessity to effect the truncation of the series and to calculate the function error for the latitude, the function error for the longitude and the function error for the azimuth. In this research, these series are truncated in the derivative third and calculates the express functions error in derivative fourth. It is described the adjustment models based on the least-squares method: combined model with weighted parameters, combined model or mixed model, parametric model or observations equations and correlates model or condition equations model. The practical application is the adjustment by mean parametric model of a traverse measured by the Instituto Brasileiro de Geografia e Estatística (IBGE), constituted of 8 vertices and the 129.661 km length. The localization of errors in the observations is calculated by the Baarda s data snooping test in the last iteration of the adjustment that showed some observations with error. The estimates of quality are in the variance-covariance matrices and calculate the semiaxes of the error ellipse or standard ellipse of each point by means of the spectral decomposition (or Jordan s decomposition) of the submatrices of the variance-covariance matrix of the adjusted parameters (the coordinates). It is important to note that the application of the Legendre s series is satisfactory for short distances until 40km length. The convergence of the series is fast for the adjusted coordinates, where the stopped criterion of the iterations is four decimals in the sexagesimal second arc, where it is obtained from interation second of the adjustment. / Ajustamento de linhas poligonais na superfície do elipsóide com os objetivos de garantir a unicidade de solução no transporte de coordenadas geodésicas curvilíneas (latitude ϕ e longitude λ ) e no transporte de azimute e de obter as estimativas de qualidade. Deduz o transporte de coordenadas e o transporte de azimute pelas séries de Legendre da linha geodésica. Essa série se fundamenta na série de Taylor, em que o argumento é o comprimento da linha geodésica. Para as aplicações práticas, há a necessidade de efetuar o truncamento da série e calcular a função erro para a latitude, função erro para a longitude e função erro para o azimute. Nesta pesquisa, trunca-se a série na derivada terceira e calculam-se as funções erro expressas em derivada quarta. Expõe os modelos de ajustamento fundamentados no método dos mínimos quadrados (MMQ): modelo combinado com ponderação aos parâmetros, modelo combinado ou implícito, modelo paramétrico ou das equações de observação e modelo dos correlatos ou das equações de condição. A aplicação prática é o ajustamento pelo modelo paramétrico de uma linha poligonal medida pelo Instituto Brasileiro de Geografia e Estatística (IBGE), constituída de 8 vértices e de comprimento igual a 129,661 km. A localização de erros nas observações é efetuada pelo teste data snooping de Baarda na última etapa do ajustamento que mostrou algumas observações com erro. As estimativas de qualidade estão nas matrizes variância-covariância (MVC) e calcula-se os semieixos da elipse dos erros (ou elipse padrão) de cada ponto mediante a decomposição espectral (ou decomposição de Jordan) das submatrizes da MVC dos parâmetros (as coordenadas) ajustados. Mostra-se que a aplicação das séries de Legendre é satisfatória para distâncias curtas até 40km. A convergência da série é rápida para as coordenadas ajustadas, onde o critério de parada das iterações seja quatro decimais do segundo de arco em que se atingiu na segunda etapa do ajustamento.