Robot vision with cellular neural networks: a practical implementation of new algorithms

Cellular neural networks (CNNs) are well suited for image processing due to the possibility of a parallel computation. In this paper, we present two algorithms for tracking and obstacle avoidance using CNNs. Furthermore, we show the implementation of an autonomous robot guided using only real‐time visual feedback; the image processing is performed entirely by a CNN system embedded in a digital signal processor (DSP). We successfully tested the two algorithms on this robot. Copyright © 2006 John Wiley & Sons, Ltd.     

The impact of European bioinformatics

Europe is undergoing a political integration process, the goals of which are still under discussion. Some countries, such as Germany, France, Italy, Spain, Holland, and Belgium, have coordinated aspects of their economy, including a common currency, under the administration of the European Commission. Key countries such as the UK and Denmark are still undecided about their degree of participation, and others, including most of the former Eastern-block countries, are actively seeking integration. The EC has limited power in equilibrium with the European parliament and national and regional governments. Direct funding of the sciences was not included in the EC's mandate. It was considered a strategic responsibility of the national governments. This political concept evolved into new modes of collaboration between countries and the EC. Bioinformatics could be a key area for these developments, because, if properly coordinated, it has great potential for generating added value to the new genomics and proteomic technology. Currently, some bioinformatics-related research is funded as networks of groups from separate countries, and others are funded as basic research services, such as databases. For example, the E-Biosci project, coordinated by the European Molecular Biology Laboratory as a large research institution, includes the participation of groups from different countries and one or two medium-sized companies. Other computational science and telecommunications-related programs are rarely accessible to the bioinformatics community.     

Flexural properties of sawdust reinforced phenolic composites: Pilot study

The advantageous properties of sawdust and phenolic resins were combined by making sawdust reinforced phenolic composites with different percentages by weight of sawdust. The sawdust was divided into three grades in accordance with its particulate size. Garamite and propylene glycol were added individually and together to enhance the flexural properties of the composites obtained. Without any garamite and propylene glycol, it was discovered that the best flexural properties of the composites were obtained when the percentage by weight of sawdust (<300 μm) is up to 15%. Beyond this, the flexural properties dropped significantly; in addition, the fluidity of the composite was very low and the mixture was not suitable for casting. In general, the flexural modulus of the composites decreases with an increase in sawdust content, i.e., they are more elastic but their maximum flexural strain does not improve. Garamite was therefore added to improve the maximum flexural strains of the composites and this was successful. The addition of propylene glycol makes the composite more plastic. © 2009 Wiley Periodicals, Inc. J Appl Polym Sci, 2009     

Uncertainty contribution of tip-sample angle to AFM lateral measurements

AFM measurements are very important for quality control in the photovoltaic, microfluidic, electronic or micro-optic industries. This work proposes an algorithm to complete the uncertainty evaluation of AFM systems along the XY-axis under conditions where tolerance of curved surfaces must be controlled. This algorithm is also tested for tilt angles between tip and sample from 0° to 9° using an experimental arrangement which consists of an AFM instrumented with an inclinometer and four step height standards.Results show good agreement between the theoretical model and experimental results for samples with larger steps TGZ03 (465 nm) and TGZ11 (1416 nm), but with poor results for the smaller samples TGZ01 (17.6 nm) and TGZ02 (73.1 nm). An angle of 9° shows an error of about 3% in the horizontal determination of the step dimension, but it could increase to 47% for a tilt angle of 30° according to the theoretical model.The angle error between tip and sample is included in the uncertainty budget using a uniform distribution. An evaluation is performed in a theoretical rolling machine for imprint lithography where a step must be measured with nominal dimensions of 3 μm—X-axis and 1 μm—Z-axis. An assumed tip-sample angle is assumed that changes from 0° to 22.5° (curved form) and produces an uncertainty contribution to the X measurement of 55.7 nm. This uncertainty is important and must be considered to guarantee tolerances in quality control of curved form products.