Clifford Algebra based neuralnetworks offer an alternative model to the more traditional architectures allowing: a concise abstract modelling of complex multidimensional problem domains;reduced network size compared ...
详细信息
Clifford Algebra based neuralnetworks offer an alternative model to the more traditional architectures allowing: a concise abstract modelling of complex multidimensional problem domains;reduced network size compared with traditional network configurations;improved training algorithm efficiency compared with conventional training algorithms.
Recent progress in supervised image classification research, has demonstrated the potential usefulness of incorporating fuzziness in the training, allocation and testing stages of several classification techniques. In...
详细信息
Recent progress in supervised image classification research, has demonstrated the potential usefulness of incorporating fuzziness in the training, allocation and testing stages of several classification techniques. In this paper a multiresolution neural network approach to supervised classification is presented, exploiting the inherent fuzziness of such techniques in order to perform classification at different resolution levels and gain in computational complexity. In particular, multiresolution image analysis is carried out and hierarchical neuralnetworks are used as an efficient architecture for classification of the derived multiresolution image representations. A new scheme is then introduced for transferring classification results to higher resolutions based on the fuzziness of the results of lower resolutions, resulting in faster implementation. Experimental results on land cover mapping applications from remotely sensed data illustrate significant improvements in classification speed without deterioration of representation accuracy.
Human beings perceive images through their properties, like colour, shape, size, and texture. Texture is a fertile source of information about the physical environment. images of low density crowds tend to present coa...
详细信息
Human beings perceive images through their properties, like colour, shape, size, and texture. Texture is a fertile source of information about the physical environment. images of low density crowds tend to present coarse textures, while images of dense crowds tend to present fine textures. This paper describes a new technique for automatic estimation of crowd density, which is a part of the problem of automatic crowd monitoring, using texture information based on grey-level transition probabilities on digitized images. Crowd density feature vectors are extracted from such images and used by a self organizing neural network which is responsible for the crowd density estimation. Results obtained respectively to the estimation of the number of people in a specific area of Liverpool Street Railway Station in London (UK) are presented.
Although bar-codes are normally thought of in the context of the labelling of products in retail situations, they increasingly find use in a manufacturing context for identification purposes. This paper describes a te...
详细信息
Although bar-codes are normally thought of in the context of the labelling of products in retail situations, they increasingly find use in a manufacturing context for identification purposes. This paper describes a technique which locates bar-codes in a computer image of a manufacturing cell. The bar code detecting sub-system is based on a neural network. Two neural network paradigms are compared for use in this application. The work is part of a programme of research in which the aim is to develop a system which locates and decode bar-codes in real-time.
In object recognition using neuralnetworks the correct selection of features is essential for achieving successful generalization of a net as well as satisfying time performance during the training and recognition ph...
详细信息
In object recognition using neuralnetworks the correct selection of features is essential for achieving successful generalization of a net as well as satisfying time performance during the training and recognition phase. This paper shows possibilities of automatically supporting this task in two steps. In the first step, a given feature set is examined with respect to its class separating capabilities. In the second step, the feature set is stripped of redundancies using the input pruning method, which is applied to trained networks. The experiments carried out to data show that the feature set valuation and feature selection procedure described above lead to a meaningful reduction of the feature space and thus automatically support a time consuming part of the experimental phase of feature selection before classification.
It has been shown that, for limited databases, neuralnetworks can be used effectively to identify which spectra, from a set of template spectra, are present in a set of test spectra. Identification can be achieved ev...
详细信息
It has been shown that, for limited databases, neuralnetworks can be used effectively to identify which spectra, from a set of template spectra, are present in a set of test spectra. Identification can be achieved even in the presence of large amounts of noise. It has also been shown that the linear effects of weighting and offset can have serious effects on the performance of neuralnetworks. Normalization, prior to training and testing, is the only remedy for such effects. Spectral combinations can also be accurately identified by neural network if the network is trained on examples of combinations. Singular value decomposition, which was developed as a general linear least-squares fitting algorithm, has been shown to be very effective in this study. It was shown that for linear effects, SVD can give virtually perfect results using the full 160-spectrum database. Noise was shown to be a problem but the iterative process devised by the authors greatly improve matters.
The authors introduce the conventional multilayer perceptron and Vapnik's (1995) theory of local estimation: results comparing these techniques are given. They then present two applications of prediction: one in a...
详细信息
The authors introduce the conventional multilayer perceptron and Vapnik's (1995) theory of local estimation: results comparing these techniques are given. They then present two applications of prediction: one in a check processing center, and the other in a credit card call center. Both applications are presently operating at sites run by SLIGOS. Finally, they briefly describe the ACE ESPRIT project, which is aimed at integrating neuralnetworks into a software platform for prediction in trading rooms.
Phase-measuring profilometry is a technique for reconstructing the three-dimensional shape of the surface of an object from a periodic, usually sinusoidally varying, structured light pattern generated on the object. T...
详细信息
Phase-measuring profilometry is a technique for reconstructing the three-dimensional shape of the surface of an object from a periodic, usually sinusoidally varying, structured light pattern generated on the object. This paper describes preliminary work on the use of neuralnetworks to identify phase wraps earlier in the phase measuring process, prior to the calculation of wrapped phase. The results of this work suggest that phase images can be unwrapped using the methods we have described. The neural network was capable reliably detecting phase wraps in noisy data. However, the reliability of the phase-wrap detection deteriorates as more noise is added and no substantial advantage over existing methods has yet been demonstrated. A more careful selection of neural network inputs and the application of more advanced training algorithms is likely to improve the generalisation ability of the neural network to detect the presence of phase discontinuities.
The high potential for applications of neural nets was realized in the corporate research divisions of Siemens in Munich and Princeton 10 years ago. The broad range of applications on which the central research divisi...
详细信息
The high potential for applications of neural nets was realized in the corporate research divisions of Siemens in Munich and Princeton 10 years ago. The broad range of applications on which the central research division cooperates with the business units is illustrated. Furthermore. a sampling of representative applications is given. The application spectrum can be categorized into several main classes which are modeling, optimization, forecast, decision support, control and diagnosis. The specific applications range from single large scale solutions (e.g. steel production and processing) to solutions with an appreciable multiplication factor (e.g. pulp production, water purification) and solutions for mass products (e.g. decoding of TV signals).
Throughout Europe beef carcasses are classified by visual inspection according to the so-called EUROP scheme regulated by the European Union. The EUROP scheme defines fatness and conformation (shape) on scales from 1 ...
详细信息
Throughout Europe beef carcasses are classified by visual inspection according to the so-called EUROP scheme regulated by the European Union. The EUROP scheme defines fatness and conformation (shape) on scales from 1 to 15. This is the basis for payment to the farmer. The classification is carried out by the meat plant's classifier-a person trained to grade according to the EUROP scheme. The classifiers are controlled by government inspectors, which in turn are controlled by a board of European inspectors. The paper describes a computer vision system using neuralnetworks which is able to outperform and replace the classifier.
暂无评论