Ingeniería en Sistemas, Electrónica e Industrial
Permanent URI for this communityhttp://repositorio.uta.edu.ec/handle/123456789/1
Browse
15 results
Search Results
Item Sistema de detección de emociones mediante el análisis de indicadores faciales empleando inteligencia artificial(Universidad Técnica de Ambato. Facultad de Ingeniería en Sistemas, Electrónica e Industrial. Carrera de Telecomunicaciones, 2024-02) Fiallos Valladares, Daniel Rodrigo; Córdova Córdova, Edgar PatricioEmotions are essential in several areas of life, however, their detection and understanding can become complicated, this will lead to misunderstandings and hinder communication, negatively impacting people's social relationships. In this context, the research was carried out with the aim of implementing an emotion detection system by analyzing facial indicators using artificial intelligence and visualizing the emotions that a person may have for a period of time. The system is divided into three stages, starting with the acquisition and processing of data through the activation and use of a webcam, supported by the OpenCV library for image processing techniques. The training phase involves the development of a deep learning model using Convolutional Neural Networks from facial recognition using FaceNet, perfected its design through data fitting, the architecture of the neural network focused on the extraction and learning of relevant features. Finally, the storage and visualization stage, the data is processed by the Jetson Nano and sent to a web hosting environment that receives the results and transmits them to the administrative interface for the management and visualization of the user's emotion report. The test results indicated that the system captured frames every 4 seconds, and boasts a classification accuracy of 92%, considering that the model has an outstanding ability to classify emotions in real time.Item Sistema de reconocimiento de indicadores de somnolencia mediante inteligencia artificial(Universidad Técnica de Ambato. Facultad de Ingeniería en Sistemas, Electrónica e Industrial. Carrera de Telecomunicaciones, 2023-09) Altamirano Guerra, Mayra Dennise; Córdova Córdova, Edgar PatricioThe lack of sleep not only affects safety but also increases the risk of other health problems. Sleepiness, mainly caused by sleep deprivation, negatively impacts daily human functions, including reaction time, performance, and attention, leading to a decrease in alertness and concentration. In this context, a study was conducted with the aim of implementing an artificial intelligence-based system to recognize signs of sleepiness and issue alerts to individuals in that state, in order to restore their attention and allow them to continue with their activities. The system consists of four stages: acquisition, processing, training, and visualization. In the acquisition stage, the Pi Noir V2 camera was used to capture real-time images or videos. The acquired data was sent to the NVIDIA Jetson Nano for processing. Neural networks were used to train a model capable of accurately recognizing indicators of sleepiness. For practical use and deployment, the system was implemented in a cloud hosting environment. The system's algorithm was developed in Python due to the variety of available libraries, and the OpenCV library was used for image processing due to its wide range of commands. Test results showed that the system processes and sends information at an average time of 2.38 milliseconds for real-time video.Item Sistema automático para el control de la calidad del calzado mediante visión artificial(Universidad Técnica de Ambato. Facultad de Ingeniería en Sistemas, Electrónica e Industrial. Carrera de Telecomunicaciones, 2023-03) Laura Nata, Ana Gabriela; Jurado Lozada, Marco AntonioThe objective of the current work was to implement an automatic system for the quality control of footwear. It pretends to provide the industrial sector with a tool to improve the process to evaluate the quality, identifying the existence of defective shoes through the detection of failures using Deep Learning and Computer Vision algorithms. To continue, the implementation of the automatic quality control system for footwear starts from the selection of the electronic components to be part of the project, considering the hardware and software requirements that help the compatibility between them. Subsequent, the prototype has a conveyor belt that is responsible for moving the footwear to the artificial vision booth. This compartment has an ultrasonic sensor that detects if there is a product inside it and sends the signal to the Arduino to stop the band for an estimated time of 20 seconds. Then, the four cameras capture photos and detect any problem, later, save the results to the database. In fact, the development of the system presents a detection about of footwear defect types such as threads, paint, and glue using the YOLOv5 model, which is trained through a process that is responsible for learning the neural network. Finally, the results are presented through evaluation parameters of the failure detection system through confusion matrices and validation of the results of the network training, obtaining an accuracy of 94.7%. In addition, regarding the quality of the footwear, there is an average detection of 83.80% of coincidences in the recognition.Item Pronóstico de precipitación en el centro de Tungurahua aplicando aprendizaje estadístico con redes neuronales artificiales(Universidad Técnica de Ambato. Facultad de Ingeniería en Sistemas, Electrónica e Industrial. Maestría en matemática Aplicada, 2023) López Parra, Juan Luis; Benalcázar Palacios, Marco EnriqueEste trabajo presenta el análisis de aplicación de redes neuronales artificiales recurrentes para el pronóstico de precipitación en la estación meteorológica de Querochaca, provincia de Tungurahua. Los datos analizados pertenecen a la lluvia del periodo de 2015 al 2021. El modelo propuesto consiste en una red neuronal artificial recurrente de 5 capas. La primera capa es la celda de memoria LSTM, que se destaca por tomar en cuenta la secuencialidad de los datos que son objeto de análisis. Además, se utilizaron 4 capas completamente conectadas , donde la primera trabaja con función de activación tanh y 100 neuronas. Las dos capas que siguen a continuación tienen función de activación relu con 500 neuronas cada una. Estas tres capas mencionadas, trabajaron con factor dropout de 0,02. La capa de salida tiene 1 neurona con función de activación purelin. La arquitectura general de la red neuronal consiste en una primera capa con celda de memoria LSTM, seguida de tres capas ocultas y finalmente la capa de salida. La red neuronal ha sido configurada para desarrollar regresión lineal. Se realizaron pruebas con 10, 50, 100, 500 y 1000 neuronas respectivamente en la segunda, tercera y cuarta capa. Además, se trabajó con 10 y 100 unidades ocultas en la celda LSTM. Se seleccionó la mejor arquitectura en base a dos indicadores de rendimiento que son el error RMSE y la gráfica de pronóstico. Los hiperparámetros de entrenamiento que más influyen sobre el pronóstico son el valor máximo de épocas y la tasa de aprendizaje inicial. En este proyecto se obtuvieron los mejores resultados utilizando pocas unidades ocultas en la celda LSTM. El mejor modelo construido en el presente trabajo presenta un error 4,4535 con una gráfica de pronóstico muy similar al comportamiento real.Item Sistema para minimizar el riesgo durante la entrega de los estudiantes de educación inicial a los representantes usando técnicas de reconocimiento facial en la Unidad Educativa Cristiana New Life,(Universidad Técnica de Ambato. Facultad de Ingeniería en Sistemas, Electrónica e Industrial. Carrera de Tecnologías de la Información, 2022-09) Calo Tisalema, Christian Fernando; Nogales Portero, Rubén EduardoThe use of facial recognition systems has been in great demand in recent years as they offer greater security during the processes carried out in companies and organizations, and are even used by government entities focused on citizen security. They are also present in educational institutions for the control of personnel, or in the case of students for the registration of their attendance. The research project aims to develop a facial recognition system that reliably identifies the representatives or persons authorized to pick up students, in order to ensure a safe delivery of early childhood education students during dismissal. The current process is carried out manually, where teachers, based on their knowledge and experience, are in charge of delivering the students. The system was created in the Matlab development environment in its R2020a version, MySQL version 8.0 was used for the database manager and the facial recognition technique consisted in the application of Neural Networks, with the Alexnet pretrained network. The development process was carried out through the application of the agile methodology Extreme Programming (XP).Item Algoritmos de procesamiento de señales para el reconocimiento facial y de voz empleando redes neuronales(Universidad Técnica de Ambato. Facultad de Ingeniería en Sistemas, Electrónica e Industrial. Carrera de Ingeniería en Electrónica y Comunicaciones, 2022-09) Orozco Analuiza, Carlos Alexander; Pallo Noroña, Juan PabloThe present titling work deals with the development of an access control system through facial and voice recognition, for the authentication of people in a home. Currently, the levels of insecurity have increased and this has led to an increase in theft and damage to real estate due to the low level of security in a home. The device developed in this project is based on a bimodal access biometric, through machine learning and neural networks, a subfield of Artificial Intelligence. The system has two authentication methods: facial and voice, for which neural network models designed by the researcher were used with the stages of: database formation, image and audio processing, and neural network design. He implements two methods of facial and voice authentication to avoid identity theft, a camera is responsible for capturing the person's face and performing recognition through the neural network, if the user is registered, the microphone is activated to capture the key of access and process it through the neural network, to record the data a LAMP server is used where the system information and user notifications are stored through the Telegram application. This project is aimed at controlling the access of people to a home, avoiding the use of traditional authentication methods.Item Evaluación con red neuronal del proceso de corte láser por CO 2 en materiales compuestos de fibra de cabuya(Universidad Técnica de Ambato. Facultad de Ingeniería en Sistemas, Electrónica e Industrial. Maestría en Física Aplicada, 2022) Almache Barahona, Verónica Carolina; Pérez Salinas, Cristian FabiánThe application of Machine Learning today has allowed the development of learning models to solve problems in different fields of industry. This research work focused on relating neural networks (ANN) with the manufacture of composite materials (polyester matrix + fiber cabuya) and CO2 laser cutting machining. The objective is to develop a neural network to evaluate the application of machine learning to predict the surface finish characteristic of the material under study. The established cutting parameters were laser power and cutting speed. The surface finish characteristic to be evaluated was the surface roughness of the cut composite material. The sheet of the constructed composite material was subjected to CO2 laser cutting, which generated a set of 84 specimens. Experimental data was generated by measuring surface roughness through laboratory tests. The programming of the neural network was done with the Scikit-learn package. This is one of the most widely used open source libraries for machine learning available in Python. The results achieved by the prediction of the network based on the experimental data are related to the values predicted by the neural network model (ANN) and the performance of the network was evaluated using statistical metrics. The statistical results obtained were 0.946, 0.139 and 0.301 corresponding to the coefficient of determination (R2), the mean square error (MSE), and the mean absolute error (MAE) respectively. Therefore, it could be concluded that the performance of the developed neural network has a high validity and ability to predict surface roughness.Item Desarrollo de una aplicación de realidad aumentada y visión artificial para el mantenimiento de ventiladores mecánicos(Universidad Técnica de Ambato. Facultad de Ingeniería en Sistemas, Electrónica e Industrial. Maestría en Electrónica y Automatización, 2022) Poveda Ocaña, Héctor Fabián; Galarza Zambrano, Eddie EgbertoThe present work shows the development of a maintenance system for the Savina 300 intensive care ventilator using augmented reality as a guide for the user. The identification algorithm has used a neural network trained with a 3D model of the respirator and the configuration of several relevant details for the training process. The objective execution platform is a mobile device with Android 11 operating system and motion sensors (gyroscope and accelerometer) that allows versatility, mobility and a low-cost implementation of the solution. The detection has been successful in execution times which allows an overlay of the 3D elements on top of the real fan. The proposed structure and work route offers robustness for the implementation of more maintenance tasks in the future. The validation has been carried out using the well-known SUS usability test in which the application reached a value of 82.5, which indicates that it is advisable to make improvements at the user interface and message display level in case of needing a better coefficient.Item Sistema de control de acceso por medio de reconocimiento facial con uso de mascarilla y monitoreo de temperatura(Universidad Técnica de Ambato. Facultad de Ingeniería en Sistemas, Electrónica e Industrial. Carrera de Ingeniería en Electrónica y Comunicaciones, 2022-03) Untuña Toalombo, Verónica de los Ángeles; Altamirano Meléndez, SantiagoThe world is going through a health crisis that has affected many people, the number of infections has been gradually increasing due to non-compliance with biosafety regulations by citizens. The present degree work describes the development of a control system through facial recognition with the use of a mask and temperature monitoring, where there is control of the acquired data, which are stored for later analysis. The control system is made up of three stages, these being: data acquisition through the temperature sensor and a webcam to later be sent to the Raspberry Pi development card so that it receives the information and performs the process of recognition of use. of masks through convolutional neural networks, which detect whether or not a person is wearing the mask, the information is sent through the MQTT network protocol that transports messages between devices, a mobile app is used to view the data which is linked to ThingSpeak, which is an Internet of things platform, in case the temperature exceeds its nominal value, an alert signal will be sent to the Telegram messaging application. This project is aimed at controlling the entry of people, whether in public or private places, avoiding direct contact with control personnel, and monitoring the correct temperature for the respective entry.Item Implementación de un sistema inteligente para la identificación vehicular(Universidad Técnica de Ambato. Facultad de Ingeniería en Sistemas, Electrónica e Industrial. Maestría en Matemática Aplicada, 2021) Cáceres Mayorga, Paúl Alejandro; Reinoso Astudillo, Cristina IsabelThe main objective of this research work was to implement an intelligent system capable of classifying vehicle and automotive license plates, as well as self-correcting recognition errors, for which it was based on the design of an algorithm capable of detecting and classifying vehicles, implementing artificial intelligence. Once the process to be followed was identified, a source code was implemented for the detection of the types of license plates using the convolutional neural network WPOD in which the data of the edge, width and height of the plate were specified so that it only provides the photo of license plate. For the binarization process used in the research, the Otsu algorithm was used, which converts the images into the gray, blur, binary and dilation scales, applying filters that can obtain the location segments of the letters and numbers. Finally, an effective system was obtained, with acceptable detection capacity, since design parameters of the architecture of each type of red were established, which achieved a satisfactory solution to the problem of identification, classification and validation of the characters of Ecuadorian license plates.