Search

The bachelor thesis is the implementation of an idea for the license plate recognition system which using neural network. This project contains two main parts; The first one is the preprocessing image and the other part is the training neural network for license plate recognition from those inputs from pre-processing process. As the preprocessing image part has been done in the previous senior project, I have taken the result from the senior project to develop the algorithm for the high accuracy plate localizing and digits segmentation. Subsequently, the neural network is built based on the dataset of modified handwritten digits, which was provided by the founder of MI-AI community, then the preprocessed output from the preprocessing image stage will be fitted and the system makes the prediction based on the pretrain dataset. The system aims to recognize the Vietnamese Auto License Plate from images which captured randomly of cars in Viet Nam, or pictures were captured from “Black-box-cam" or “Dashcam” on cars as database. The application of this system will support for Auto-Parking, AccessControl in specified area/region, and Journey Time Measurement, etc

The bachelor thesis is the implementation of an idea for the license plate recognition system which using neural network. This project contains two main parts; The first one is the preprocessing image and the other part is the training neural network for license plate recognition from those inputs from pre-processing process. As the preprocessing image part has been done in the previous senior project, I have taken the result from the senior project to develop the algorithm for the high accuracy plate localizing and digits segmentation. Subsequently, the neural network is built based on the dataset of modified handwritten digits, which was provided by the founder of MI-AI community, then the preprocessed output from the preprocessing image stage will be fitted and the system makes the prediction based on the pretrain dataset. The system aims to recognize the Vietnamese Auto License Plate from images which captured randomly of cars in Viet Nam, or pictures were captured from “Black-box-cam" or “Dashcam” on cars as database. The application of this system will support for Auto-Parking, AccessControl in specified area/region, and Journey Time Measurement, etc

The bachelor thesis is the implementation of an idea for the license plate recognition system which using neural network. This project contains two main parts; The first one is the preprocessing image and the other part is the training neural network for license plate recognition from those inputs from pre-processing process. As the preprocessing image part has been done in the previous senior project, I have taken the result from the senior project to develop the algorithm for the high accuracy plate localizing and digits segmentation. Subsequently, the neural network is built based on the dataset of modified handwritten digits, which was provided by the founder of MI-AI community, then the preprocessed output from the preprocessing image stage will be fitted and the system makes the prediction based on the pretrain dataset. The system aims to recognize the Vietnamese Auto License Plate from images which captured randomly of cars in Viet Nam, or pictures were captured from “Black-box-cam" or “Dashcam” on cars as database. The application of this system will support for Auto-Parking, AccessControl in specified area/region, and Journey Time Measurement, etc

Приведен анализ методов разреженного представления и преобразования цифровых изображений для экстракции и уменьшения исходного пространства признаков с целью поиска и распознавания лиц. Представлены сравнительные вычислительные сложности ряда методов, The article presents the analysis of methods for sparse representation and converting the digital image, extraction and reduce the original feature space search and recognition. Considered comparative computational complexities of these methods: DCT, FFT, DWT, PCA, NNDA

A large amount of data are input into the computer. These data can be presented as visual images, sounds (speech), electronic signals, etc. And there is a problem with their recognition. Recognition of graphic symbol, including letters and numbers is the most common one.There are three basic methods of graphic symbols identification: reference, structural and feature ones. For example, programs like FineReader, Readiris, ScanSoft, OmniPage, CuneiForm can be used to recognize the graphic symbols, which are entered into the computer. Significant reliance recognition accuracy of image quality and some distortion of primary image are main disadvantages of these programs and methodsWe can use fuzzy logic or neural network algorithms to improve recognition accuracy.For example neural networks are used in Google technology.We offered the use of numbers recognition system based on fuzzy logic to improve the accuracy and the feature method was proposed.As informative parameters, i.e. recognition features, two positive and two negative transitions were chosen. Transition is defined as the difference between the current local maximum (or minimum) and the previous local minimum (or maximum) number of pixels (intensity) which include images. The number of pixels is counted within vertical sectors into which the number of images is divided and the width of these sectors may vary from 1 to 3 pixels. Another informative parameter is the number of transitions counted in the separation of the image number into the horizontal sectors.Using of Fuzzy Logic Designer in LabVIEW for design and realization of the fuzzy logic system is considered in the article. The recognition system was designed as a software component in package LabVIEW.The research of the developed recognition system in terms of distortion of numbers images by their rotation was held. The angle of this turn may vary from 0 to 10 both clockwise and counterclockwise.By turning clockwise the range of rotation angle is 2-4, В этой работе рассмотрена разработка и исследование системы распознавания цифр на основе нечеткой логики в среде LabVIEWДля увеличения точности предложено использовать систему распознавания цифр на основе нечеткой логики с использованием признакового метода распознавания.В работе рассмотрено построение системы нечеткого вывода и ее реализация с использованием Fuzzy Logic Designer среды LabVIEW.Были проведены испытания разработанной системы распознавания в условиях искажений изображений цифр путем их повороту на градус в диапазоне от 0 до 10, как за часовой стрелкой, так и против., В даній роботі розглянуто розробку та дослідження системи розпізнавання цифр на основі нечіткої логіки в середовищі LabVIEW.Для підвищення точності запропоновано використовувати систему розпізнавання цифр на базі нечіткої логіки з використанням ознакового методу розпізнавання.В роботі розглянуто побудову системи нечіткого виводу та її реалізація з використанням Fuzzy Logic Designer середовища LabVIEW.Були проведені дослідження розробленої системи розпізнавання в умовах спотворення зображень цифр шляхом їх повороту на кут в діапазоні від 0 до 10, як за годинниковою стрілкою так і проти.

Discovered and proved the overall structure of the algorithms used in pattern recognition. A comparative analysis of methods and approaches to solving the problem of finding and identifying human face on the image. A modified algorithm for image scaling and clustering, which reduces the number of informative tracts candidates for the location of the research object (human face) is proposed. It is established that the use of neural networks coagulation makes a small number of errors in a large number of coagulation and other layers. It is established that the network has a large invariance to position the face in the picture. In consequence of that generalization ability higher than the multilayer perceptron. Evaluating the effectiveness of probabilistic systems showed that the use of the proposed approaches and algorithms enables a high probability likely face recognition (93%). Results of the study can be used in the development of automated systems for access: a personal computer, a bank account, application to access data on the image of a human face in miniature devices where there is no possibility to embed common hardware identification.

This master thesis project is carried out by one student at the Royal Institute of Technology in collaboration with Cybercom Group. The aim was to evaluate and compare system design strategies for fruit recognition in harvesting robots and the performance of supervised machine learning classification methods when applied to this specific task. The thesis covers the basics of these systems; to which parameters, constraints, requirements, and design decisions have been investigated. The framework is used as a foundation for the implementation of both sensing system, and processing and classification algorithms. A plastic tomato plant with fruit of varying maturity was used as a basis for training and testing, and a Kinect v2 for Windows including sensors for high resolution color-, depth, and IR data was used for image acquisition. The obtained data were processed and features of objects of interest extracted using MATLAB and a SDK for Kinect provided by Microsoft. Multiple views of the plant were acquired by having the plant rotate on a platform controlled by a stepper motor and an Ardunio Uno. The algorithms tested were binary classifiers, including Support Vector Machine, Decision Tree, and k-Nearest Neighbor. The models were trained and validated using a five fold cross validation in MATLABs Classification Learner application. Peformance metrics such as precision, recall, and the F1-score, used for accuracy comparison, were calculated. The statistical models k-NN and SVM achieved the best scores. The method considered most promising for fruit recognition purposes was the SVM., Det här masterexamensarbetet har utförts av en student från Kungliga Tekniska Högskolan i samarbete med Cybercom Group. Målet var att utvärdera och jämföra designstrategier för igenkänning av frukt i en skörderobot och prestandan av klassificerande maskininlärningsalgoritmer när de appliceras på det specifika problemet. Arbetet omfattar grunderna av dessa system; till vilket parametrar, begränsningar, krav och designbeslut har undersökts. Ramverket användes sedan som grund för implementationen av sensorsystemet, processerings- och klassifikationsalgoritmerna. En tomatplanta i pplast med frukter av varierande mognasgrad användes som bas för träning och validering av systemet, och en Kinect för Windows v2 utrustad med sensorer för högupplöst färg, djup, och infraröd data anvöndes för att erhålla bilder. Datan processerades i MATLAB med hjälp av mjukvaruutvecklingskit för Kinect tillhandahållandet av Windows, i syfte att extrahera egenskaper ifrån objekt på bilderna. Multipla vyer erhölls genom att låta tomatplantan rotera på en plattform, driven av en stegmotor Arduino Uno. De binära klassifikationsalgoritmer som testades var Support Vector MAchine, Decision Tree och k-Nearest Neighbor. Modellerna tränades och valideras med hjälp av en five fold cross validation i MATLABs Classification Learner applikation. Prestationsindikatorer som precision, återkallelse och F1- poäng beräknades för de olika modellerna. Resultatet visade bland annat att statiska modeller som k-NN och SVM presterade bättre för det givna problemet, och att den sistnömnda är mest lovande för framtida applikationer.

This master thesis project is carried out by one student at the Royal Institute of Technology in collaboration with Cybercom Group. The aim was to evaluate and compare system design strategies for fruit recognition in harvesting robots and the performance of supervised machine learning classification methods when applied to this specific task. The thesis covers the basics of these systems; to which parameters, constraints, requirements, and design decisions have been investigated. The framework is used as a foundation for the implementation of both sensing system, and processing and classification algorithms. A plastic tomato plant with fruit of varying maturity was used as a basis for training and testing, and a Kinect v2 for Windows including sensors for high resolution color-, depth, and IR data was used for image acquisition. The obtained data were processed and features of objects of interest extracted using MATLAB and a SDK for Kinect provided by Microsoft. Multiple views of the plant were acquired by having the plant rotate on a platform controlled by a stepper motor and an Ardunio Uno. The algorithms tested were binary classifiers, including Support Vector Machine, Decision Tree, and k-Nearest Neighbor. The models were trained and validated using a five fold cross validation in MATLABs Classification Learner application. Peformance metrics such as precision, recall, and the F1-score, used for accuracy comparison, were calculated. The statistical models k-NN and SVM achieved the best scores. The method considered most promising for fruit recognition purposes was the SVM., Det här masterexamensarbetet har utförts av en student från Kungliga Tekniska Högskolan i samarbete med Cybercom Group. Målet var att utvärdera och jämföra designstrategier för igenkänning av frukt i en skörderobot och prestandan av klassificerande maskininlärningsalgoritmer när de appliceras på det specifika problemet. Arbetet omfattar grunderna av dessa system; till vilket parametrar, begränsningar, krav och designbeslut har undersökts. Ramverket användes sedan som grund för implementationen av sensorsystemet, processerings- och klassifikationsalgoritmerna. En tomatplanta i pplast med frukter av varierande mognasgrad användes som bas för träning och validering av systemet, och en Kinect för Windows v2 utrustad med sensorer för högupplöst färg, djup, och infraröd data anvöndes för att erhålla bilder. Datan processerades i MATLAB med hjälp av mjukvaruutvecklingskit för Kinect tillhandahållandet av Windows, i syfte att extrahera egenskaper ifrån objekt på bilderna. Multipla vyer erhölls genom att låta tomatplantan rotera på en plattform, driven av en stegmotor Arduino Uno. De binära klassifikationsalgoritmer som testades var Support Vector MAchine, Decision Tree och k-Nearest Neighbor. Modellerna tränades och valideras med hjälp av en five fold cross validation i MATLABs Classification Learner applikation. Prestationsindikatorer som precision, återkallelse och F1- poäng beräknades för de olika modellerna. Resultatet visade bland annat att statiska modeller som k-NN och SVM presterade bättre för det givna problemet, och att den sistnömnda är mest lovande för framtida applikationer.

Les instal·lacions del clavegueram són d’una importància extrema per mantenir una correcta situació d’higiene en una ciutat. Un cop construïdes, és necessari fer-ne un manteniment regular per garantir que segueixen complint la seva funció, però degut a la seva naturalesa subterrània i de conductes foscos i estrets, aquesta no sempre és una tasca senzilla. L’objectiu d’aquest projecte és contribuir al desenvolupament d’una eina que permeti realitzar aquests manteniments d’una forma més senzilla, segura i eficient. Aquesta eina està sent desenvolupada pel departament d’ESAII de la UPC, per petició de l’empresa de Fomento y Contratas (FCC), els encarregats actualment de fer el manteniment del clavegueram de Barcelona. La seva finalitat principal serà l’exploració d’embornals, que tot i ser la part del clavegueram que més a prop es troba de la superfície, no resulten fàcils d’obrir i inspeccionar, motiu pel qual es vol un aparell que pugui realitzar el procés d’exploració de forma directa i sense necessitat d’aixecar-ne la reixa. El projecte consistirà en desenvolupar una part del software d’aquesta eina. Concretament, es centra al voltant del disseny d’un algorisme que permeti convertir un conjunt de punts capturats en un espai en 3D i que representen l’interior d’aquest embornal a una dada numèrica que indicarà quin percentatge de la canonada que connecta amb el sistema de clavegueram principal està obstruïda, perquè l’equip de manteniment pugui valorar si és necessari accedir a l’embornal per fer una neteja o no., Las instalaciones de alcantarillado son de una importancia extrema para mantener una correcta situación de higiene en una ciudad. Una vez construidas, es necesario realizarles un mantenimiento regular para garantizar que siguen cumpliendo su función, pero debido a su naturaleza subterránea y de conductos oscuros y estrecho, ésta no siempre resulta tarea fácil. El objetivo de este proyecto es contribuir al desarrollo de una herramienta que permita realizar estos mantenimientos de una forma mas sencilla, segura y eficiente. Esta herramienta esta siendo desarrollada por el departamento de ESAII de la UPC, por petición de la empresa Fomento y Contratas (FCC), los actuales encargados de hacer el mantenimiento del alcantarillado de Barcelona. Su finalidad principal será la exploración de imbornales, que aunque són la parte del alcantarillado más cercana a la superfície, no resultan fáciles de abrir e inspeccionar, motivo por el qual se desea un aparato que sea capaç de realizar el proceso de exploración de forma directa y sin necesidad de levantar la reja. El proyecto consistirá en desarrollar una parte del software de esta herramienta. Concretamente, se centra alrededor del diseño de un algoritmo que permita convertir un conjunto de puntos capturados en un espacio de 3D y que representan el interior de este imbornal en un dato numérico que indicará que porcentage de la cañería que conecta con el sistema principal de alcantarillado está obstruída, para que el equipo de mantenimiento pueda valorar si es necesario acceder al interior para realizar una limpieza o no., In any city, sewer installations are of major importance to help maintain a correct hygienic state. Once built, it’s necessary to do regular maintenance on them to guarantee they keep accomplishing their function properly. Provided they are built in an underground System filled with dark and small conduits, this is not usually an easy task. The aim of this project is to contribute to the development of a tool that aids in doing this maintenance in an easier, safer and more eficient way. This tool is being developed by the UPC’s ESAII department by petition of the Fomento and Contratas company, the corrent responsables for doing Barcelona’s city sewer maintenance. It’s main purpose is the exploration of scuppers, which even though are the most superficial part of the sewer System, are not particularly easy to open and examine manually. For this reason, the company wants a device capable of doing this exploration process directly and without even needing to lift the case. This project will be centered around the developing a part of the software for this device. In particular, we want to design an algorithm that is capable of transforming a particular set of points in a 3D space which represent the inside of a scupper into a numerical value which indicates the portion of the pipe connection to the main draining system is obstructed. This value will help the maintenance team in taking a decision regarding wether to manually access the scupper for cleaning or not.

In this article a recognition system of isolated sequences based on noise filtration by microphone array is described. The configuration of microphone array has an important issue for clarification of beamformer spatial filtration. Beamforming is a method for processing microphone array data to produce images that represent the distribution of the acoustic source strength. Recognition systems are based on classification systems pattern. In these systems, sounds or sequences of sounds, such as phonemes, groups of phonemes, or words are modeled by distinct classes. The goal of the recognition system is to estimate the correct sequence of classes, sounds, that make up the incoming utterance and hence, the sounds that were reproduced. In particular, many new distance measures between time series have been introduced. The recognition system based on time series analysis, where the Dynamic time warping algorithm based on spatial filtration preprocessing is discussed. © 2015 IEEE.

In this article a recognition system of isolated sequences based on noise filtration by microphone array is described. The configuration of microphone array has an important issue for clarification of beamformer spatial filtration. Beamforming is a method for processing microphone array data to produce images that represent the distribution of the acoustic source strength. Recognition systems are based on classification systems pattern. In these systems, sounds or sequences of sounds, such as phonemes, groups of phonemes, or words are modeled by distinct classes. The goal of the recognition system is to estimate the correct sequence of classes, sounds, that make up the incoming utterance and hence, the sounds that were reproduced. In particular, many new distance measures between time series have been introduced. The recognition system based on time series analysis, where the Dynamic time warping algorithm based on spatial filtration preprocessing is discussed. © 2015 IEEE.

RIVM rapport:Het erkenningstraject wordt breed gewaardeerd en heeft de afgelopen jaren goed gefunctioneerd: het maakt de kwaliteit van interventies voldoende inzichtelijk en het stimuleert de kwaliteitsverbetering bij de ontwikkeling van interventies. Over de betekenis in de praktijk ligt de mening over de bijdrage van het traject genuanceerder: dat het traject niet voldoende bij de praktijk aansluit is het meest geuite punt van kritiek. Concreet betekent dit dat complexe interventies niet goed in het systeem passen en dat de meerwaarde van erkenning nog onvoldoende duidelijk is. Ook geven professionals aan te weinig tijd te hebben om hun interventies voor erkenning in te dienen. Verbeterpunten zijn onder andere: stel de eisen voor onderzoek van effectiviteit bij en breng de meerwaarde van erkenning voor interventie-eigenaren sterker naar voren. Uit aanvullend onderzoek op het terrein van gezondheidsbevordering blijkt dat specifieke aandacht nodig is voor de vraag 'wat werkt voor wie onder welke omstandigheden?' Maak dus explicieter waar erkende interventies uit bestaan en wat ze veronderstellen van de lokale context. En bied een breder palet aan leeren verbeterprocessen, zoals ondersteuning van professionals, om het functioneren van de gezondheidsbevordering te verbeteren. Dit blijkt uit de evaluatie die in 2011 is uitgevoerd naar het proces en resultaat van het erkenningstraject voor interventies., The Dutch Recognition System Interventions is well appreciated and has functioned as intended and planned: the system provides more insight into the quality of interventions and triggers quality improvement in the development of new interventions. The extent in which the recognition system contributes to improving professional practice remains less clear; its lack of connection with the practice of public health and youth (social) care was a concern that was raised often. Complex interventions do not fit easy in the system, the added value of a recognition is still insufficiently understood and professionals do not have the required time to submit their intervention. Suggestions for improvement include adjustment of the requirements for effectiveness and communication of the added value of a recognition for the owners. Additional research with respect to health promoting practice, showed that a more specific question needs to be answered: what works and under which circumstances or conditions? The various elements that make up an intervention have to be identified more explicitly, as should the necessary requirements for implementation at local level. In addition, a broader package of learning and improvement processes is needed to improve health promoting practice. This was all shown in the evaluation of the process and outcomes of the Recognition System Interventions, that was conducted in 2011.

RIVM rapport:Het erkenningstraject wordt breed gewaardeerd en heeft de afgelopen jaren goed gefunctioneerd: het maakt de kwaliteit van interventies voldoende inzichtelijk en het stimuleert de kwaliteitsverbetering bij de ontwikkeling van interventies. Over de betekenis in de praktijk ligt de mening over de bijdrage van het traject genuanceerder: dat het traject niet voldoende bij de praktijk aansluit is het meest geuite punt van kritiek. Concreet betekent dit dat complexe interventies niet goed in het systeem passen en dat de meerwaarde van erkenning nog onvoldoende duidelijk is. Ook geven professionals aan te weinig tijd te hebben om hun interventies voor erkenning in te dienen. Verbeterpunten zijn onder andere: stel de eisen voor onderzoek van effectiviteit bij en breng de meerwaarde van erkenning voor interventie-eigenaren sterker naar voren. Uit aanvullend onderzoek op het terrein van gezondheidsbevordering blijkt dat specifieke aandacht nodig is voor de vraag 'wat werkt voor wie onder welke omstandigheden?' Maak dus explicieter waar erkende interventies uit bestaan en wat ze veronderstellen van de lokale context. En bied een breder palet aan leeren verbeterprocessen, zoals ondersteuning van professionals, om het functioneren van de gezondheidsbevordering te verbeteren. Dit blijkt uit de evaluatie die in 2011 is uitgevoerd naar het proces en resultaat van het erkenningstraject voor interventies., The Dutch Recognition System Interventions is well appreciated and has functioned as intended and planned: the system provides more insight into the quality of interventions and triggers quality improvement in the development of new interventions. The extent in which the recognition system contributes to improving professional practice remains less clear; its lack of connection with the practice of public health and youth (social) care was a concern that was raised often. Complex interventions do not fit easy in the system, the added value of a recognition is still insufficiently understood and professionals do not have the required time to submit their intervention. Suggestions for improvement include adjustment of the requirements for effectiveness and communication of the added value of a recognition for the owners. Additional research with respect to health promoting practice, showed that a more specific question needs to be answered: what works and under which circumstances or conditions? The various elements that make up an intervention have to be identified more explicitly, as should the necessary requirements for implementation at local level. In addition, a broader package of learning and improvement processes is needed to improve health promoting practice. This was all shown in the evaluation of the process and outcomes of the Recognition System Interventions, that was conducted in 2011.

RIVM rapport:Het erkenningstraject wordt breed gewaardeerd en heeft de afgelopen jaren goed gefunctioneerd: het maakt de kwaliteit van interventies voldoende inzichtelijk en het stimuleert de kwaliteitsverbetering bij de ontwikkeling van interventies. Over de betekenis in de praktijk ligt de mening over de bijdrage van het traject genuanceerder: dat het traject niet voldoende bij de praktijk aansluit is het meest geuite punt van kritiek. Concreet betekent dit dat complexe interventies niet goed in het systeem passen en dat de meerwaarde van erkenning nog onvoldoende duidelijk is. Ook geven professionals aan te weinig tijd te hebben om hun interventies voor erkenning in te dienen. Verbeterpunten zijn onder andere: stel de eisen voor onderzoek van effectiviteit bij en breng de meerwaarde van erkenning voor interventie-eigenaren sterker naar voren. Uit aanvullend onderzoek op het terrein van gezondheidsbevordering blijkt dat specifieke aandacht nodig is voor de vraag 'wat werkt voor wie onder welke omstandigheden?' Maak dus explicieter waar erkende interventies uit bestaan en wat ze veronderstellen van de lokale context. En bied een breder palet aan leeren verbeterprocessen, zoals ondersteuning van professionals, om het functioneren van de gezondheidsbevordering te verbeteren. Dit blijkt uit de evaluatie die in 2011 is uitgevoerd naar het proces en resultaat van het erkenningstraject voor interventies., The Dutch Recognition System Interventions is well appreciated and has functioned as intended and planned: the system provides more insight into the quality of interventions and triggers quality improvement in the development of new interventions. The extent in which the recognition system contributes to improving professional practice remains less clear; its lack of connection with the practice of public health and youth (social) care was a concern that was raised often. Complex interventions do not fit easy in the system, the added value of a recognition is still insufficiently understood and professionals do not have the required time to submit their intervention. Suggestions for improvement include adjustment of the requirements for effectiveness and communication of the added value of a recognition for the owners. Additional research with respect to health promoting practice, showed that a more specific question needs to be answered: what works and under which circumstances or conditions? The various elements that make up an intervention have to be identified more explicitly, as should the necessary requirements for implementation at local level. In addition, a broader package of learning and improvement processes is needed to improve health promoting practice. This was all shown in the evaluation of the process and outcomes of the Recognition System Interventions, that was conducted in 2011.

We explore the use of natural language understanding and image processing to index and query American Football tapes. We present a model for representing spatio-temporal characteristics of multiple objects in dynamic scenes in this domain, and a recognition system which uses the model to recognise American Football plays.

This paper presents an algorithm for effective learning of the Adaptive Resonance Theory (ART-1) artificial neural network and shows the convenience of applying this network for object identification, on the basis of which the zero version of the "ART-1 Simulator" software was developed for new generation recognition systems which are used today in the control of intelligent industrial robot based on CCD camera involvement. The parameters of the ART-1 structure are in accordance with captured images and learning procedure, adopted for identification of objects in robotic environment.

All speech recognition systems, whether speaker-independent or speaker-dependent, require large amounts of training data to estimate the model parameters and, generally, the more training data available, the better the recognition performance. To improve the recognition performance of a system for a new speaker without having to train entirely new models, adapting the existing models during the recognition process is a practical solution. This report describes an investigation into the subject of recognition-time speaker adaptation of a tied-mixture HMM recognition system, with the goal of implementing a system which adapts to a new speaker during the course of its usage.

This invited paper presents an algorithm for effective machine learning of the ART-1 artificial neural network and shows the convenience of applying this network for object identification, on the basis of which the zero version of the "ART Simulator V1.0" software was developed for new generation of the recognition systems which are used today within the pattern recognition of intelligent industrial robot vision system.

This paper presents the hierarchical intelligent control of an industrial robot using a system of artificial neural networks, based on an empirical control strategy. This type of intelligent industrial robots control provides a much wider domain of autonomy in the execution of manufacturing tasks, especially manipulation of parts, so these robots have the ability of machine learning based on "own experience". Conducted research has shown that improvements in terms of programming, flexibility, efficiency and skill of an intelligent mechatronic system - industrial robot depend on the degree of development and implementation of its machine learning by using two artificial neural networks: ART-1 net for pattern recognition of object and four-layer "back-propagation" net for coordinate mapping between the CCD camera and robot coordinates of TCP.

A single Fourier transform optical system was constructed and tested on two projects. First, film grain in a single Fourier transform operation was studied and a mathematical model of the film grain and a semi-quantitative prediction of the grain's transform were developed and verified using the experimental system. The film grain was found to have a noise-like spatial frequency spectrum which was a decreasing function of frequency, and that this spectrum could be considered Gaussian only in the case of fine grain films containing binary signal images. Second, a system was used to investigate a proposed method of detecting a ship wave. A deterministic relation between a ship wave and the Fourier transform of its photograph was shown to exist. A sample photograph of a ship wave was inserted into the system and operated on by a matched filter rotated In a plane normal to the optical axis., http://archive.org/details/anexperimentalsi1094512177, Lieutenant, United States Navy, Approved for public release; distribution is unlimited.