Por favor, use este identificador para citar o enlazar este ítem:
http://dspace.udla.edu.ec/handle/33000/11041| Tipo de material : | bachelorThesis |
| Título : | Herramienta de reconocimiento facial con técnica de visión computacional 2D |
| Autor : | Escobar Loayza, Jefferson David |
| Tutor : | Hernández Perdomo, Wilmar |
| Palabras clave : | BIOMETRÍA;RECONOCIMIENTO FACIAL;VISIÓN POR COMPUTADORA;INTELIGENCIA ARTIFICIAL |
| Fecha de publicación : | 2019 |
| Editorial : | Quito: Universidad de las Américas, 2019 |
| Citación : | Escobar Loayza, J. D. (2019). Herramienta de reconocimiento facial con técnica de visión computacional 2D (Tesis de pregrado). Universidad de las Américas, Quito |
| Resumen : | En esta tesis se ha diseñado un sistema de reconocimiento de imágenes por visión artificial, centrado en rostros de personas. Aquí, se ha estudiado la complejidad de dos algoritmos que han demostrado su eficiencia y robustez en aplicaciones de reconocimiento de rostros usando visión por computador... |
| Descripción : | In this thesis an image recognition system has been designed by artificial vision, centered on people's faces. Here, we have studied the complexity of two algorithms that have demonstrated their efficiency and robustness in face recognition applications using computer vision. In addition, a system has been implemented for such purpose based on said algorithms and the proposal of improvements in them. The algorithms mentioned above are: 1) The method Autorostros (Eigenfaces) and 2) The method of Linear Discriminant Analysis (Fisherfaces). Both methods are able to detect faces by comparison with images stored in databases, and among the fundamental characteristics of these methods we can highlight the following: the method of Autorostros is responsible for comparing with a single image of the person who serves as a reference and that is stored in the database; while the Linear Discriminant Analysis method performs the comparison against an average of reference images stored in the database. In this thesis it has been possible to demonstrate experimentally that the first method needs less processing time, but is less robust; while the second method requires a longer processing time, but is more robust. Therefore, the latter is appropriate for working in environments affected by pollution and unwanted signals. With the above background, in order to improve the robustness of the method Autorostros, it is proposed to create a database with a set of reference photos of each of the faces to identify, find a measure of the central tendency of these photos to each face and take this measure as the new reference. However, even doing this, the results are still inferior to the Linear Discriminant Analysis method. Therefore, if a rapid identification system is desired and a certain margin of error is tolerated, the recommended method is the Autorostros. On the other hand, if you want to implement a robust processing system, the appropriate method is Linear Discriminant Analysis. |
| URI : | http://dspace.udla.edu.ec/handle/33000/11041 |
| Aparece en las colecciones: | Ingeniería en Electrónica y Redes de Información |
Ficheros en este ítem:
| Fichero | Descripción | Tamaño | Formato | |
|---|---|---|---|---|
| UDLA-EC-TIERI-2019-12.pdf | 4,63 MB | Adobe PDF | Visualizar/Abrir |
Este ítem está sujeto a una licencia Creative Commons Licencia Creative Commons
