The aim of this master's thesis thesis is to enable evaluation of quality, or the type of product in industrial applications using artificial neural networks, especially in applications where the classical approach of machine vision is too complicated. The system thus designed is implemented onto a specific hardware platform and becomes a subject to the final optimalisation for the hardware platform for the best performance of the system.

This master’s thesis deals with establishment of parameters of the PPI filter material using methods of image analysis. The first part is aimed at general methods used in practice, an overview of splitting into filter clasification and characterization of the PPI filter material. The second part represents an introduction to problems of machine vision. There are outlined basic methods of image processing. Last parts of the thesis study the method using detection of bright light levels. The measurement and interpretation of results is performed with the PPI material - types 10/6 and 10/12. For image processing is used the MATLAB program, version 7.0.1.

This diploma thesis deals with a practical application employing an industrial robot KUKA, a vision system – smart camera Siemens. The application is focused on synchronizing or robot movements with objects moving on a conveyor belt. The introductory and theoretical part of this thesis is concerned with various systems for machine vision currently available on the market. Practical part is then focused on the demonstration application: setting-up the robotic cell and description of all devices, robot and vision system programming.

Object inspection sensor is a device that is mainly used in the automation of the production process. The Diploma Thesis deals with the familiarization of optical inspection systems, survey and comparison of companies of these systems in the Czech Republic. ELLA-CS company and dafault state output control products in her workplace is presented in the Diploma Thesis. The next part concerns the design of the inspection system for the company ELLA-CS, which is designed on the platform Raspberry. The thesis also deals with the implementation of the inspection system, which describes the hardware and software of this system. The last part is intended to discuss the results and suggestions for improving the inspection.

This work deals with analysis one of the production lines at ModusLink Czech Republic s.r.o. It seeks to find possibilities for automation of different processes on the line, currently performed manually. The basic focus is to find the appropriate technical solutions to the automated control of the designated production line. The work seeks to describe all meaningful solutions and compare their advantages and disadvantages. From this comparison of the final proposal is based on comprehensive solutions with the use of machine vision. Another point is the selection of a suitable contractor for the implementation of the proposal, making its advantage in economic terms and the resulting recommendations, if implemented the proposal.

<p>Manufacturers have begun to adopt advanced digital methods and tools to release higher-quality products more quickly at a lower cost. Model-based technologies and Visual/Vision Inspection System (VIS) are two representative approaches and are trends to facilitate manufacturers’ digitalization. With model-based technologies, manufacturers can define digital product requirements and specifications as the primary data source for all engineering activities during the entire product lifecycle, including design, manufacturing, and inspection, rather than relying on traditional paper-based documents. Moreover, the manufacturers can improve product defect detection efficiency and reduce labor requirements by implementing a VIS without artificial visual inspection in their inspection process. Traditionally, inspectors rely heavily on information from 2D engineering drawings to inspect manufactured products with/without a VIS. This inspection process is out of the model-based context, hindering manufacturers from transiting paper-based to model-based processes.</p> <p>To answer the question of how we can adapt a VIS to support model-based inspection, the author selected a developmental research methodology to address the research gap of model-based visual inspection. An image-based Model-Based Visual Inspection System (MBVIS) framework integrating Model-Based Definition (MBD) and VIS applications was proposed to demonstrate the process of utilizing the Geometric Dimensioning and Tolerancing (GD&T) information from the MBD dataset and driving the VIS to measure physical targets. A plastic bottle with a high dimensional standard obtained from a plastic bottle manufacturer was chosen to evaluate the proposed system’s performance in extracting GD&T information from the MBD dataset and measuring relevant critical bottle dimensions. </p> <p>The interchangeability of the proposed MBVIS with a commercial VIS from the partner bottle company in measuring critical bottle dimensions was proven by conducting the Bland and Altman (B&A) analysis and other statistical analyses in this research. Moreover, two other customized tests were run to understand the effectiveness of the GD&T extraction and measurement result output processes within the MBVIS framework. The results indicated that both processes were reliable in extracting GD&T information and exporting measurement results.</p>

This paper describes methods to track a user-defined point in the vision of a robot as it drives forward. This tracking allows a robot to keep itself directed at that point while driving so that it can get to that user-defined point. I develop and present two new multi-scale algorithms for tracking arbitrary points between two frames of video, as well as through a video sequence. The multi-scale algorithms do not use the traditional pyramid image, but instead use a data structure called an integral image (also known as a summed area table). The first algorithm uses edge-detection to track the movement of the tracking point between frames of video. The second algorithm uses a modified version of the Moravec operator to track the movement of the tracking point between frames of video. Both of these algorithms can track the user-specified point very quickly. Implemented on a conventional desktop, tracking can proceed at a rate of at least 20 frames per second.

No description available.

Unmanned ground vehicles (UGVs) are increasingly being used for civilian and military applications. Passive sensing, such as visible cameras, are being used for navigation and object detection. An additional object of interest in many environments is text. Text information can supplement the autonomy of unmanned ground vehicles. Text most often appears in the environment in the form of road signs and storefront signs. Road hazard information, unmapped route detours and traffic information are available to human drivers through road signs. Premade road maps lack these traffic details, but with text localization the vehicle could fill the information gaps. Leading text localization algorithms achieve ~60% accuracy; however, practical applications are cited to require at least 80% accuracy [49]. The goal of this thesis is to test existing text localization algorithms against challenging scenes, identify the best candidate and optimize it for scenes a UGV would encounter. Promising text localization methods were tested against a custom dataset created to best represent scenes a UGV would encounter. The dataset includes road signs and storefront signs against complex background. The methods tested were adaptive thresholding, the stroke filter and the stroke width transform. A temporal tracking proof of concept was also tested. It tracked text through a series of frames in order to reduce false positives. Best results were obtained using the stroke width transform with temporal tracking which achieved an accuracy of 79%. That level of performance approaches requirements for use in practical applications. Without temporal tracking the stroke width transform yielded an accuracy of 46%. The runtime was 8.9 seconds per image, which is 44.5 times slower than necessary for real-time object tracking. Converting the MATLAB code to C++ and running the text localization on a GPU could provide the necessary speedup. / Master of Science

[ES] El objetivo de esta Tesis Doctoral es generar conocimiento y tecnología que permitan desarrollar soluciones mecanizadas y automatizadas mediante la aplicación de sensores para monitorizar, optimizar tareas y aumentar la productividad en los cultivos de cítricos y, por ende, la competitividad del sector citrícola. Con este fin, se ha trabajado en la aplicación de tecnologías en algunas de las operaciones más importantes y costosas en la citricultura. En el ámbito de la recolección, con el objetivo de facilitar y optimizar el proceso de recolección de cítricos para consumo en fresco, se implementaron mejoras en un prototipo de plataforma autopropulsada de asistencia para garantizar la seguridad de los trabajadores y minimizar los riesgos laborales, asegurar la calidad de la fruta recolectada y clasificarla adecuadamente en función de su calidad por un sistema de visión artificial. Tras la implementación de estos avances, se evaluó su funcionalidad en condiciones reales de campo y se determinó el rendimiento de la operación de recolección asistida. Los resultados mostraron que el rendimiento operativo teórico en el proceso de recolección con la asistencia del prototipo fue un 17% menor que en la recolección manual tradicional. Asimismo, con la finalidad de generar de manera automática mapas de rendimiento en cítricos accesibles al agricultor, se creó una nueva herramienta a modo de cuadro de mando, denominada CitrusYield. Esta herramienta es capaz de obtener mapas que muestran la variabilidad, haciendo uso de la información proporcionada al inspeccionar cada fruta y de los datos obtenidos por los dispositivos de geolocalización, permitiendo conocer el rendimiento de la producción a nivel intraparcelario. Se evalúo el funcionamiento de CitrusYield con datos obtenidos por el prototipo. El cuadro de mando analizó los datos y generó diferentes mapas georreferenciando la información de la fruta con su punto de recolección. A través de estos mapas, CitrusYield demostró su capacidad para reflejar la heterogeneidad de la producción dentro de la parcela. Los indicadores de calidad, rendimiento del sistema de inspección por visión artificial y la productividad revelaron el grado de eficiencia del proceso de recolección con la asistencia del prototipo. En el ámbito de la detección temprana de plagas y enfermedades, se llevó a cabo la investigación del potencial de la imagen en color y de la imagen hiperespectral para la detección automática de daños causados por la plaga Tetranychus urticae en hojas de cítricos. Además, se diferenciaron en función de la antigüedad y de otros problemas que puede tener el árbol causados por deficiencias nutricionales o por otras plagas. En el análisis para identificar el daño causado por esta plaga, se obtuvo una tasa de éxito del 92,5% en las imágenes en color. No obstante, no se pudo discriminar la edad del daño. En las imágenes hiperespectrales, el éxito fue del 100% en la detección del daño y del 92% en la discriminación de la edad. En cuanto a la discriminación de daños de T. urticae respecto a otros daños, en las imágenes en color se obtuvo una tasa de éxito del 100% en la discriminación de Phyllocnistis citrella en el envés de las hojas. Las deficiencias de N no pudieron discriminarse correctamente y las deficiencias de Fe, Mn o Zn se detectaron acertadamente en el 50% de los casos. En las imágenes hiperespectrales, los daños por P. citrella se discriminaron con éxito en el 100% de los casos. Las deficiencias por N se discriminaron correctamente en más del 65% de los casos. No obstante, las deficiencias de Fe, Mn o Zn no pudieron ser discriminadas correctamente. Las conclusiones de esta Tesis Doctoral ponen de manifiesto el gran potencial de implementar estas tecnologías para monitorizar, optimizar tareas y digitalizar los procesos en las labores llevadas a cabo en los campos de cítricos, de manera que se logra aumentar la eficiencia, productividad y, también, la competitividad en el sector. / [CA] L'objectiu d'esta Tesi Doctoral és generar coneixement i tecnologia que permeten desenvolupar solucions mecanitzades i automatitzades mitjançant l'aplicació de sensors per a monitoritzar, optimitzar tasques i augmentar la productivitat en els cultius de cítrics i, per tant, la competitivitat del sector citrícola. Amb este fi, s'ha treballat en l'aplicació de tecnologies en algunes de les operacions més importants i costoses de la citricultura. Amb l'objectiu de facilitar i optimitzar el procés de recol·lecció de cítrics per a consum en fresc es van implementar millores en un prototip de plataforma autopropulsada d'assistència per a garantir la seguretat dels treballadors i minimitzar els riscos laborals, assegurar la qualitat de la fruita recol·lectada i classificar-la adequadament en funció de la qualitat estimada pel sistema de visió artificial. Després de la implementació d'estos avanços, es va avaluar la seua funcionalitat en condicions reals de camp i es va determinar el rendiment de l'operació de recol·lecció assistida. Els resultats van mostrar que el rendiment operatiu teòric en el procés de recol·lecció amb l'assistència del prototip va ser un 17% menor que en la recol·lecció manual tradicional. Així mateix, amb la finalitat de generar de manera automàtica mapes de rendiment en cítrics accessibles a l'agricultor, es va crear una nova eina digital a manera de quadre de comandament, denominada CitrusYield. Esta eina és capaç d'obtindre mapes que mostren la variabilitat, fent ús de la informació proporcionada pel sistema de visió artificial i de les dades obtingudes pels dispositius de geolocalización, permetent conéixer el rendiment de la producció a nivell intraparcel·lari. El quadre de comandament va analitzar les dades i va generar diferents mapes georreferenciant l'informació de la fruita amb el seu punt de recol·lecció. A través d'estos mapes, CitrusYield va demostrar la seua capacitat per a reflectir l'heterogeneïtat de la producció dins de la parcel·la. Els indicadors de qualitat, rendiment del sistema d'inspecció per visió artificial i la productivitat calculats per CitrusYield van revelar el grau d'eficiència del procés de recol·lecció amb l'assistència del prototip. En l'àmbit de la detecció de plagues i malalties de cítrics, es va dur a terme la investigació del potencial de la imatge en color i de la imatge hiperespectral, per a la detecció automàtica de danys causats per la plaga Tetranychus urticae en fulls de cítrics. A més, es van discriminar en funció de l'antiguitat i d'altres problemes que pot tindre l'arbre causats per deficiències nutricionals o per altres plagues. En l'anàlisi per a identificar el dany causat per esta plaga, es va obtindre una taxa d'èxit del 92,5% en les imatges en color. No obstant això, no es va poder discriminar l'edat del dany. En les imatges hiperespectrals, l'èxit va ser del 100% en la detecció del dany i del 92% en la discriminació de l'edat. En quan a la discriminació de danys de T. urticae respecte a altres danys, en les imatges en color es va obtindre una taxa d'èxit del 100% en la discriminació de Phyllocnistis citrella en el revés dels fulls. Les deficiències de N no van poder discriminar-se correctament i les deficiències de Fe, Mn o Zn es van detectar encertadament en el 50% dels casos. En les imatges hiperespectrals, els danys per pogué P. citrella es van discriminar amb èxit en el 100% dels casos, tant en el feix com en el revés dels fulls. Les deficiències per N es van discriminar correctament en més del 65% dels casos també en ambdós costats del full. No obstant això, les deficiències de Fe, Mn o Zn no van poder ser discriminades correctament Les conclusions de esta Tesi Doctoral posen de manifest el gran potencial d'implementar estes tecnologies per a monitoritzar, optimitzar tasques i digitalitzar els processos en les labors dues a terme en els camps de cítrics, de manera que s'aconseguix augmentar l'eficiència, productivitat i la competitivitat en el sector. / [EN] This PhD dissertation aims to generate knowledge and technology to develop mechanised and automated solutions to monitor processes, optimise tasks and increase the productivity of citrus crops and, therefore, the competitiveness of the citriculture sector. For this purpose, this thesis has focused on the development of technologies in some of the most important and costly operations in the citrus sector. To facilitate and optimise the harvesting of citrus fruits for the fresh market, improvements were implemented in a prototype of a self-propelled assistance platform. The research has focused on guaranteeing the operator safety, ensuring the quality of the harvested fruit and classifying it properly based on its quality by an artificial vision system in real-time. The functionality was evaluated in-field conditions and performance was assessed. The results showed that the assisted harvesting process with the platform was more efficient than traditional manual harvesting. A theoretical operation yield was obtained with the prototype, 17% lower than in conventional manual harvesting. Furthermore, in order to automatically generate citrus yield maps accessible for the farmer, a new digital tool was created as a dashboard, called CitrusYield. This tool is capable of obtaining maps showing the variability, using the information provided during the fruit inspection at harvest and the data obtained by the geolocation device, and also allowing to know the performance of the production at the intra-orchard level. The CitrusYield functionality was evaluated with data obtained by the prototype. The dashboard analysed the data and generated different maps georeferencing the information of the fruit with its harvesting point. Through these maps, the ability of CitrusYield to show the heterogeneity of production within the orchard was demonstrated. The production variability was also quantified in the tables and graphs generated by the tool. The indicators of fruit quality obtained by the artificial vision system and the productivity revealed the degree of effectiveness and efficiency of the harvesting process with the prototype. In the context of early citrus pests and diseases detection, the potential of colour and hyperspectral imaging, as a fast and automatic methods to detect the presence and the age of the damage caused by the Tetranychus urticae pest on citrus leaves, were investigated. Another goal was to discriminate this damage from that caused by other pests and from nutritional deficiencies. In the analysis of identifying the damage caused by T. urticae, a success rate of 92,5% was obtained in the colour images. However, the age of the damage could not be discriminated. In hyperspectral images, the success rate was 100% in damage detection and 92% in age discrimination. Regarding the discrimination of T. urticae damages from others, in the colour images, a success of 100% was obtained in the discrimination of Phyllocnistis citrella on the underside of the leaves. N deficiencies could not be correctly discriminated and Fe, Mn or Zn deficiencies were correctly detected in 50% of cases. In the hyperspectral images, P. citrella damage was successfully discriminated in 100% of cases, both on the upper and under sides of the leaves. N deficiencies were correctly discriminated in more than 65% of cases on both sides of the leaf. However, Fe, Mn or Zn deficiencies could not be correctly discriminated. The conclusions of this Thesis show the potential of implementing these technologies to monitor, optimize tasks and digitize agricultural processes in the operations usually done in citrus crops. Therefore, the efficiency, productivity and competitiveness of the sector can increase when these technologies are used. / Esta Tesis Doctoral ha recibido una subvención para la formación de personal investigador en agroalimentación en los centros de investigación agraria y alimentaria INIA-CC.AA. (FPI-INIA CPD2016-0007, #42), otorgada por el Instituto Nacional de Investigaciones y Tecnología Agraria y Alimentaria, con el apoyo parcial de los Fondos FEDER 2014-2020. En esta Tesis Doctoral se han realizado investigaciones en el marco de varios proyectos de investigación: Aplicación de nuevas tecnologías para una estrategia integral de la recolección mecanizada de cítricos (CITRUSREC) (RTA2014-00025-C05-00).Subproyecto 1 centrado en el desarrollo y evaluación de una plataforma de asistencia a la recolección (RTA2014-00025-C05-01) Subproyecto 5 sobre la generación de herramientas de ayuda para la toma de decisiones (RTA2014-00025-C05-05) Desarrollos de ingeniería para el aseguramiento de una agricultura rentable, sostenible y competitiva del campo a la mesa (GVA-IVIA proyecto 51918) Grupo Operativo "Avances tecnológicos para la modernización y la sostenibilidad en la producción de cítricos (GO CITRUSTECH)". / González González, MG. (2022). Estudio y desarrollo de soluciones mecanizadas automatizadas mediante la aplicación de sensores y nuevas tecnologías para la producción de cítricos [Tesis doctoral]. Universitat Politècnica de València. https://doi.org/10.4995/Thesis/10251/186174