Object segmentation and labeling by learning from examples

We propose a system that employs low-level image segmentation followed by color and two-dimensional (2-D) shape matching to automatically group those low-level segments into objects based on their similarity to a set of example object templates presented by the user. A hierarchical content tree data structure is used for each database image to store matching combinations of low-level regions as objects. The system automatically initializes the content tree with only "elementary nodes" representing homogeneous low-level regions. The "learning" phase refers to labeling of combinations of low-level regions that have resulted in successful color and/or 2-D shape matches with the example template(s). These combinations are labeled as "object nodes" in the hierarchical content tree. Once learning is performed, the speed of second-time retrieval of learned objects in the database increases significantly. The learning step can be performed off-line provided that example objects are given in the form of user interest profiles. Experimental results are presented to demonstrate the effectiveness of the proposed system with hierarchical content tree representation and learning by color and 2-D shape matching on collections of car and face images.     

Gamma-irradiation-induced Ag/SiO2 composite films and their optical absorption properties

Small Ag particles or clusters dispersed mesoporous SiO₂ composite films were prepared by a new method: First the matrix SiO₂ films were prepared by sol-gel process combined with the dip-coating technique, then they were soaked in AgNO₃ solutions followed by irradiation of γ-ray at room temperature and in ambient pressure. The structures of these films were examined by X-ray diffraction (XRD), high-resolution transmission electron microscope (HRTEM), and optical absorption spectroscopy. It has been shown that the Ag particles grown within the porous SiO₂ films are very small, and they are isolated and dispersed from each other with very narrow size distributions. With increasing the soaking concentration and an additional annealing, an opposite peakshift effect of the surface plasmon resonance (SPR) was observed in the optical absorption measurements.     

An evaluation of biological and abiotic controls for grapevine powdery mildew. 2. Vineyard trials

Grapevine powdery mildew ( Erysiphe necator) affects grape yield and fruit quality worldwide. Managers of conventional vineyards rely mainly on synthetic fungicides and sulfur to control powdery mildew, while in organic vineyards sulfur is the main control agent, often in rotation with canola-based oils, bicarbonates and biological control agents. The efficacy of those materials has not been evaluated critically under field conditions in Australia. Accordingly, a range of materials showing most promise in previous greenhouse trials (Crisp et al. 2006 Australian Journal of Grape and Wine Research 12 , pp. 192–202) were assessed via field trials in commercial vineyards. Applications of either milk or whey (alone, or mixed with a canola oil-based product), as well as applications of potassium bicarbonate (commercial formulation), all reduced the severity of powdery mildew compared with untreated vines. Eight applications of a 1:10 dilution of milk, 45 g/L whey powder or programs comprising rotations of potassium bicarbonate plus oil and whey, applied at 10–14 day intervals, reduced the severity of powdery mildew to levels not significantly different from that on vines sprayed with sulfur (wettable powder, 3–6 g/L). However, the relative control of powdery mildew by the test materials in field trials was dependent on the susceptibility of the grapevine cultivar and the extent of spray coverage achieved. In vineyards where highly susceptible cultivars were planted, and spray coverage was compromised, the resultant control of powdery mildew was reduced; and sometimes to commercially unacceptable levels.     

Modeling of a constant Voltage transformer

This paper presents a comprehensive nonlinear model of the controlled constant voltage transformer also known as the ferroresonant transformer. Saturation is a normal mode of operation for this device. This paper derives an equivalent electrical circuit that relates to the physical structure of a typical design. The level of detail includes winding resistances, continuously nonlinear magnetizing inductances, tapped windings, and leakage inductances. The paper describes methods to extract the winding resistances, leakage inductances, and hysteresis loops of the transformer and how to fit the latter into single-valued nonlinear functions. The paper compares computer simulation results of the model with those obtained analytically and experimentally. The results show that the derived circuit will be very useful for designers of the ferroresonant transformer, which is used in uninterruptible power supplies.     

High-resolution photonics-based interference suppression filter with wide passband

A new topology for a photonic signal processor, which overcomes the basic recursive frequency response problem that limits the passband range, is presented. The structure is based on a new multiple-wavelength offset-cavity structure that is cascaded with a series of unbalanced delay line structures. This not only can synthesize a very narrow notch response with good shape factor but also permits a multifold extension of the free spectral range (FSR) and passband width. Results on the interference mitigation filter demonstrate a stopband of 1% of center frequency and a fourfold increase in the FSR and passband width, while also having a very small shape factor, in excellent agreement with predictions.     

Combined RLS-LMS initialization for per tone equalizers in DMT-receivers

In discrete multitone receivers, the classical equalizer structure consists of a (real) time domain equalizer (TEQ) combined with complex one-tap frequency domain equalizers. An alternative receiver is based on a per tone equalization (PTEQ), which optimizes the signal-to-noise ratio (SNR) on each tone separately and, hence, the total bitrate. In this paper, a new initialization scheme for the PTEQ is introduced, based on a combination of least mean squares (LMS) and recursive least squares (RLS) adaptive filtering. It is shown that the proposed method has only slightly slower convergence than full square-root RLS (SR-RLS) while complexity as well as memory cost are reduced considerably. Hence, in terms of complexity and convergence speed, the proposed algorithm is in between LMS and RLS.     

Low Temperature Physics at the Laboratoire National des Champs Magnétiques Pulsés in Toulouse

An overview over past and present activities and future developments at the Toulouse pulsed magnetic field facility is given, both as far as technical developments of the infrastructure, as well as low temperature physics performed at the LNCMP are concerned.     

Microstrip stepped impedance resonator bandpass filter with an extended optimal rejection bandwidth

Bandpass filters with an optimal rejection bandwidth are designed using parallel-coupled stepped impedance resonators (SIRs). The fundamental (f/sub o/) and higher order resonant harmonics of an SIR are analyzed against the length ratio of the high-Z and low-Z segments. It is found that an optimal length ratio can be obtained for each high-Z to low-Z impedance ratio to maximize the upper rejection bandwidth. A tapped-line input/output structure is exploited to create two extra transmission zeros in the stopband. The singly loaded Q(Q/sub si/) of a tapped SIR is derived. With the aid of Q/sub si/, the two zeros can be independently tuned over a wide frequency range. When the positions of the two zeros are purposely located at the two leading higher order harmonics, the upper rejection band can be greatly extended. Chebyshev bandpass filters with spurious resonances up to 4.4f/sub o/, 6.5f/sub o/, and 8.2f/sub o/ are fabricated and measured to demonstrate the idea.     

Steady-state operation of a high-voltage multiresonant converter in a high-temperature environment

Often scientific instruments on exploratory satellites are used to collect cosmic particles. These instruments, such as the Faraday cup, require a high-voltage regulated power source. On the proposed solar probe satellite there will be instruments which will collect cosmic particles close to the sun. Naturally temperature effects on the voltage regulator are a concern. Specifically power dissipation within the regulator must stay within stringent guidelines. The high-voltage regulator circuit is a resonant flyback topology with a dissipative series switch whose on resistance controls the output voltage level. The extreme temperatures in space will change the inductance of the flyback transformer which will affect the output voltage. State space analysis techniques are used to determine how the value of the transformer inductance and capacitance in the flyback converter affects the output voltage and consequently the power dissipated by the transformer and the switch. The accuracy of the analysis is verified by comparing results with those obtained from an experimental circuit. The results can then be used to help the designer chose nominal capacitance and inductance values that will work over the temperature range of interest.     

Classification of audio radar signals using radial basis function neural networks

Radial basis function (RBF) neural networks are used to classify real-life audio radar signals that are collected by a ground surveillance radar mounted on a tank. Currently, a human operator is required to operate the radar system to discern among signals bouncing off tanks, vehicles, planes, and so on. The objective of this project is to investigate the possibility of using a neural network to perform this target recognition task, with the aim of reducing the number of personnel required in a tank. Different signal classification methods in the neural net literature are considered. The first method employs a linear autoregressive (AR) model to extract linear features of the audio data, and then perform classification on these features, i.e, the AR coefficients. AR coefficient estimations based on least squares and higher order statistics are considered in this study. The second approach uses nonlinear predictors to model the audio data and then classifies the signals according to the prediction errors. The real-life audio radar data set used here was collected by an AN/PPS-15 ground surveillance radar and consists of 13 different target classes, which include men marching, a man walking, airplanes, a man crawling, and boats, etc. It is found that each classification method has some classes which are difficult to classify. Overall, the AR feature extraction approach is most effective and has a correct classification rate of 88% for the training data and 67% for data not used for training.