The proceedings contain 79 papers. The topics discussed include: a consequence-finding approach for feature recognition in CAPP;case-based block division planning in shipbuilding;self-organizing compensating informati...
ISBN:
(纸本)2884491287
The proceedings contain 79 papers. The topics discussed include: a consequence-finding approach for feature recognition in CAPP;case-based block division planning in shipbuilding;self-organizing compensating information scheduler for computer integrated manufacturing;an environment for self-testing of logic programs;quality control of software specifications written in natural language;a user-centric methodology for building usable expert systems;fingerprint image compression by a clustering learning network;power system voltage instability monitoring with artificialneuralnetworks;an algebraic method to evaluate spatial stability in imageprocessing neuro chips;the use of neural network to predict welding parameters;and principled modeling and automatic classification for enhancing the reusability of problem-solving methods of expert systems.
The proceedings contains 144 papers on pattern recognition systems. Some of the topics discussed are pattern recognition, algorithms, symbol encoding, imageprocessing, feature extraction, edge detection, image qualit...
详细信息
The proceedings contains 144 papers on pattern recognition systems. Some of the topics discussed are pattern recognition, algorithms, symbol encoding, imageprocessing, feature extraction, edge detection, image quality, mathematical models, computer vision, Markov processes, biological applications, object recognition, fuzzy sets, learning systems, neuralnetworks, character recognition and artificial intelligence.
PREENS - a Parallel Research Execution Environment for neural Systems - is a distributed neurosimulator, targeted on networks of workstations and transputer systems. As current applications of neuralnetworks often co...
详细信息
This paper describes an image-based inspection system for surface defects, implemented on a transputer system with a high- performance transfer bus to provide fast access to large blocks of data. Data-level parallelis...
详细信息
ISBN:
(纸本)0819414786
This paper describes an image-based inspection system for surface defects, implemented on a transputer system with a high- performance transfer bus to provide fast access to large blocks of data. Data-level parallelism (the image pixel data is partitioned horizontally into slices) and task-level parallelism (the algorithms themselves can be parallelized) are utilized. Surface defects and anomalies are detected by a texture-based segmentation procedure. In the training phase the objects of interest are marked and all feature vectors implemented in the system are computed. The system uses simple statistical features, features calculated from the cooccurrence matrix, features based on texture spectrum and the fractal dimension. The time critical run phase is realized on a parallel computer. The implementation is done fully in software, allowing flexibility in the use of features and window sizes for pixel descriptors and freedom in the use of various classifier-methods. The processing speed of the segmentation system is easily scalable with the number of processing units. The performance of the system is demonstrated on an industrial visual inspection task, the recognition of surface defects of aluminium cast workpieces, where a connectionist classifier is used for pixel classification.
The proceedings contains 64 papers. Some of the topics discussed are: control systems;vision;measurements;speed control;diesel engines;systems stability;rocket motors;induction motors;neuralnetworks;fuzzy sets;learni...
详细信息
The proceedings contains 64 papers. Some of the topics discussed are: control systems;vision;measurements;speed control;diesel engines;systems stability;rocket motors;induction motors;neuralnetworks;fuzzy sets;learning systems;artificial intelligence;Petri nets;signal processing;quantum mechanical systems;parameter estimation;identification;recursive algorithms;robotics;intelligent manipulators;pattern recognition;example-based feature extraction;voice recognition;precision measurement;image distortion;atomic microscopes;bio-systems;electrocardiography;electromyography;cardiology;cardiovascular system;echocardiogram;industrial applications;multistage diagnosis;and stock prices forecasting.
Near-simultaneous, multispectral, coregistered imagery of ground target and background signatures were collected over a full diurnal cycle in visible, infrared, and ultraviolet spectrally filtered wavebands using Batt...
详细信息
ISBN:
(纸本)0819415189
Near-simultaneous, multispectral, coregistered imagery of ground target and background signatures were collected over a full diurnal cycle in visible, infrared, and ultraviolet spectrally filtered wavebands using Battelle's portable sensor suite. The imagery data were processed using classical statistical algorithms, artificialneuralnetworks and data clustering techniques to classify objects in the imaged scenes. imagery collected at different times throughout the day were employed to verify algorithm robustness with respect to temporal variations of spectral signatures. In addition, several multispectral sensor fusion medical imaging applications were explored including imaging of subcutaneous vasculature, retinal angiography, and endoscopic cholecystectomy. Work is also being performed to advance the state of the art using differential absorption lidar as an active remote sensing technique for spectrally detecting, identifying, and tracking hazardous emissions. These investigations support a wide variety of multispectral signature discrimination applications including the concepts of automated target search, landing zone detection, enhanced medical imaging, and chemical/biological agent tracking.
We present a new methodology for describing the functioning of artificial neurons, including new, as yet untested, types of behaviour. It also provides the possibility of defining artificial neurons of any order, and ...
详细信息
We present a new methodology for describing the functioning of artificial neurons, including new, as yet untested, types of behaviour. It also provides the possibility of defining artificial neurons of any order, and a wide range of functions from which to choose. As an illustration of the new formulation, a practical realization is analyzed, consisting of a multilayered neural network applied to the imageprocessing of black and white scenes, making manifest the possibilities of this new type of neuron in the field of cellular logic but with new types of processing. This is just an early stage in the development of the new neurons, so that many of their possible applications have yet to be initiated. Among them, one can already foresee those related to fuzzy models, analogue models, and many others. For these applications, it will no longer be necessary to make any change in the network design, just to make a choice from the proposed library of functions.< >
In the classification of chromosomes, as well as in many other applications, patterns must be recognized regardless of the orientation of the image. This study investigates the performance of two neuralnetworks, LVQ ...
详细信息
In the classification of chromosomes, as well as in many other applications, patterns must be recognized regardless of the orientation of the image. This study investigates the performance of two neuralnetworks, LVQ and ARTMAP, that are based on a combination of clustering and supervised learning. Each network was trained with images of several chromosome pairs, presented in a few different rotations, and tested with the training patterns at various intermediate rotations. Successful recognition was achieved over a wide range of rotation angles. Comparisons of the two nets for this application also illustrate the characteristics that influence their suitability for a variety of imageprocessing problems.
A straightforward method for implementing a neural network (NN) solution in the field of imageprocessing, for attitude and position determination, is presented in this paper. The proposed evolutive training algorithm...
详细信息
A straightforward method for implementing a neural network (NN) solution in the field of imageprocessing, for attitude and position determination, is presented in this paper. The proposed evolutive training algorithm is capable of setting the appropriate dimension of the neural network and the adequate weights interconnecting the neurons. The recommended solution is based on simulation results.< >
This paper studies neural data fusion (NDF) from the viewpoint of histogram features. Recurrent and multiple layer perceptron neuralnetworks are trained and tested for comparison and contrast. The fusion performance ...
详细信息
This paper studies neural data fusion (NDF) from the viewpoint of histogram features. Recurrent and multiple layer perceptron neuralnetworks are trained and tested for comparison and contrast. The fusion performance as seen by the reduction of the histogram variance is improved by neural network processing as well as by the design of the neural network architecture. Signal filtering and image restoration find applications in NDF. The performance of NDF under different signal-to-noise ratio conditions is studied, outperforming weighted average data fusion by 7.62 to 10.8% with a multiple layer perceptron and by 3.2% to 22.5% with a recurrent neural network.< >
暂无评论