Publication: Desarrollo de una aplicación para caracterización de rocas mediante Inteligencia Artificial
Loading...
Official URL
Full text at PDC
Publication Date
2021-06
Advisors (or tutors)
López Andrés, Sol
Editors
Journal Title
Journal ISSN
Volume Title
Publisher
Citations
Exportar
Abstract
El problema de la identificación y caracterización de las rocas que llevan a cabo los colectivos que trabajan con muestras geológicas como son los arquitectos, arqueólogos, historiadores etc. ha sido siempre un problema de importancia y de gran dificultad. Para el reconocimiento de minerales y rocas existen distintas técnicas que son aplicadas en diversos campos y que tienen diferentes tasas de éxito. El tipo de análisis que se lleva a cabo depende en gran medida de la posibilidad de llevar la muestra a un laboratorio o si por el contrario, la identificación debe realizarse in situ. Una de las técnicas más utilizada es la que está basada en la identificación mediante microscopía óptica. Para ello se precisa un microscopio petrográfico, y a partir de imágenes de láminas delgadas de las rocas, es posible determinar la composición mineralógica y así clasificar el tipo de roca. En el caso de las rocas utilizadas en la construcción de utensilios de uso común antiguos, como son las piedras de afilar, es importante determinar la hipotética cantera de origen. Para ello, es necesario el uso de otras técnicas complementarias como son la difracción de rayos X (DRX) y la fluorescencia de rayos X (FRX), entre otras. Este trabajo de fin de grado tiene como principal objetivo la creación de una herramienta informática de identificación de rocas, basada en un sistema de inteligencia artificial y asequible para colectivos no geológicos. El sistema es capaz de, a partir de imágenes obtenidas con un microscopio petrográfico de diferentes canteras, asignarlas una posible cantera origen entre las que previamente han sido analizadas. Se trata de un sistema que permite ser usado tanto en dispositivos móviles, tablets y smartphones como en ordenadores. Para ello se ha recopilado un conjunto de muestras de estas canteras ‘históricas’ y de piedras de afilar con las que han sido entrenados los algoritmos de clasificación del sistema basados en redes neuronales.
The problem of identification and characterization of rocks carried out by groups working with geological samples such as architects, archaeologists, historians, etc. has always been an important and very difficult problem. For the Recognition of minerals and rocks there are different techniques that are applied in various fields and that have different success rates. The type of analysis that is carried out depends to a great extent on the possibility of taking the sample to a laboratory or if, on the contrary, the identification must be done in situ. One of the most used techniques is based on identification by optical microscopy. This requires a petrographic microscope, and from images of thin slices of the rocks, it is possible to determine the mineralogical composition and thus classify the type of rock. In the case of rocks used in the construction of commonly used ancient tools, such as whetstones, it is important to determine the hypothetical quarry of origin. For this, it is necessary to use other complementary techniques such as X-ray diffraction (XRD) and X-ray fluorescence (XRF), among others. The main objective of this final degree project is the creation of a computer tool for rocks identification, based on an artificial intelligence system and accesible to non-geological groups. The system is able, from images obtained with a petrographic microscope of different quarries, to assign them a possible quarry origin among those that have been previously analyzed. It is a system that can be used on mobile devices, tablets and smartphones as well as computers. For this purpose, a set of samples of these 'historical' quarries and sharpening stones have been collected and used to train the classification algorithms of the system based on neural networks.
The problem of identification and characterization of rocks carried out by groups working with geological samples such as architects, archaeologists, historians, etc. has always been an important and very difficult problem. For the Recognition of minerals and rocks there are different techniques that are applied in various fields and that have different success rates. The type of analysis that is carried out depends to a great extent on the possibility of taking the sample to a laboratory or if, on the contrary, the identification must be done in situ. One of the most used techniques is based on identification by optical microscopy. This requires a petrographic microscope, and from images of thin slices of the rocks, it is possible to determine the mineralogical composition and thus classify the type of rock. In the case of rocks used in the construction of commonly used ancient tools, such as whetstones, it is important to determine the hypothetical quarry of origin. For this, it is necessary to use other complementary techniques such as X-ray diffraction (XRD) and X-ray fluorescence (XRF), among others. The main objective of this final degree project is the creation of a computer tool for rocks identification, based on an artificial intelligence system and accesible to non-geological groups. The system is able, from images obtained with a petrographic microscope of different quarries, to assign them a possible quarry origin among those that have been previously analyzed. It is a system that can be used on mobile devices, tablets and smartphones as well as computers. For this purpose, a set of samples of these 'historical' quarries and sharpening stones have been collected and used to train the classification algorithms of the system based on neural networks.
Description
Trabajo de Fin de Grado en Ingeniería de Computadores, Facultad de Informática UCM, Departamento de Arquitectura de Computadores e Inteligencia Artificial, Curso 2022-21.