Modelización del cálculo de Indicadores Ambientales, a partir de sensores ópticos de alta resolución temporal a bordo de satélite

La presente tesis doctoral tiene por objeto el estudio y análisis de técnicas y modelos de obtención de parámetros biofísicos e indicadores ambientales, de manera automatizada a partir de imágenes procedentes de satélite de alta resolución temporal. En primer lugar se revisan los diferentes programas espaciales de observación del territorio, con especial atención a los que proporcionan dicha resolución. También se han revisado las metodologías y procesos que permiten la obtención de diferentes parámetros cuantitativos y documentos cualitativos, relacionados con diversos aspectos de las cubiertas terrestres, atendiendo a su adaptabilidad a las particularidades de los datos. En segundo lugar se propone un modelo de obtención de parámetros ambientales, que integra información proveniente de sensores espaciales y de otras fuentes auxiliares utilizando, en cierta medida, las metodologías presentadas en apartados anteriores y optimizando algunas de las referidas o proponiendo otras nuevas, de manera que se permita dicha obtención de manera eficiente, a partir de los datos disponibles y de forma sistemática. Tras esta revisión de metodologías y propuesta del modelo, se ha procedido a la realización de experimentos, con la finalidad de comprobar su comportamiento en diferentes casos prácticos, depurar los flujos de datos y procesos, así como establecer las situaciones que pueden afectar a los resultados. De todo ello se deducirá la evaluación del referido modelo. Los sensores considerados en este trabajo han sido MODIS, de alta resolución temporal y Thematic Mapper (TM), de media resolución espacial, por tratarse de instrumentos de referencia en la realización de estudios ambientales. También por la duración de sus correspondientes misiones de registro de datos, lo que permite realizar estudios de evolución temporal de ciertos parámetros biofísicos, durante amplios periodos de tiempo. Así mismo. es de destacar que la continuidad de los correspondientes programas parece estar asegurada. Entre los experimentos realizados, se ha ensayado una metodología para la integración de datos procedentes de ambos sensores. También se ha analizado un método de interpolación temporal que permite obtener imágenes sintéticas con la resolución espacial de TM (30 m) y la temporal de MODIS (1 día), ampliando el rango de aplicación de este último sensor. Asimismo, se han analizado algunos de los factores que afectan a los datos registrados, tal como la geometría de la toma de los mismos y los episodios de precipitación, los cuales alteran los resultados obtenidos. Por otro lado, se ha comprobado la validez del modelo propuesto en el estudio de fenómenos ambientales dinámicos, en concreto la contaminación orgánica de aguas embalsadas. Finalmente, se ha demostrado un buen comportamiento del modelo en todos los casos ensayados, así como su flexibilidad, lo que le permite adaptarse a nuevos orígenes de datos, o nuevas metodologías de cálculo. Abstract This thesis aims to the study and analysis of techniques and models, in order to obtain biophysical parameters and environmental indicators in an automated way, using high temporal resolution satellite data. Firstly we have reviewed the main Earth Observation Programs, paying attention to those that provide high temporal resolution. Also have reviewed the methodologies and process flow diagrams in order to obtain quantitative parameters and qualitative documents, relating to various aspects of land cover, according to their adaptability to the peculiarities of the data. In the next stage, a model which allows obtaining environmental parameters, has been proposed. This structure integrates information from space sensors and ancillary data sources, using the methodologies presented in previous sections that permits the parameters calculation in an efficient and automated way. After this review of methodologies and the proposal of the model, we proceeded to carry out experiments, in order to check the behavior of the structure in real situations. From this, we derive the accuracy of the model. The sensors used in this work have been MODIS, which is a high temporal resolution sensor, and Thematic Mapper (TM), which is a medium spatial resolution instrument. This choice was motivated because they are reference sensors in environmental studies, as well as for the duration of their corresponding missions of data logging, and whose continuity seems assured. Among the experiments, we tested a methodology that allows the integration of data from cited sensors, we discussed a proposal for a temporal interpolation method for obtaining synthetic images with spatial resolution of TM (30 m) and temporal of MODIS (1 day), extending the application range of this one. Furthermore, we have analyzed some of the factors that affect the recorded data, such as the relative position of the satellite with the ground point, and the rainfall events, which alter the obtained results. On the other hand, we have proven the validity of the proposed model in the study of the organic contamination in inland water bodies. Finally, we have demonstrated a good performance of the proposed model in all cases tested, as well as its flexibility and adaptability.

Confección de cartografía temática de coberturas del suelo mediante imágenes aerotransportadas de alta resolución.

El objetivo del presente Proyecto Fin de Carrera es la elaboración de cartografía base de la zona Rivas - Vaciamadrid, situada al noreste de Madrid, a partir de imágenes de alta resolución espacial pancromáticas y en color obtenidas mediante teledetección aerotransportada de la zona. Se pretende poder facilitar el reconocimiento de la morfología y la geología natural de la zona desde la clasificación de la cobertura del suelo. La zona de trabajo actualmente está construida y en el momento del registro de datos se encontraba en estado natural. La finalidad consiste en proporcionar una información temática que permita llevar a cabo estudios de análisis de cobertura y de cambios. Se trata de una imagen en alta resolución por un sensor aerotransportado, ATM (Airbone Thematic Mapper), de naturaleza pasiva. La imagen fue registrada en el año 1997 y contiene información clasificada en 11 bandas del espectro electromagnético. El proyecto consta de dos partes: 1. Confección de cartografía base: o Documentación previa de los aspectos físicos globales (geomorfológicos, geológicos, hidrológicos) del área de estudio, a través de los documentos que puedan existir en internet con acceso libre. o Obtención de cartografía a escala 1/25000. 2. Confección de la cartografía temática: o Selección de la zona de estudio dentro de la imagen registrada y tratada en la primera parte del proyecto. o Clasificación de la imagen para análisis y definición de la cobertura del suelo. o Edición de la cartografía temática. El resultado del proyecto es una cartografía base, a escala 1/25000, que contiene información descriptiva sobre la distinta cobertura de suelo de la zona a tratar, antes de que ésta fuera construida y/o modificada artificialmente, y cartografía temática de la zona de interés.

Analysis of Thematic Classified Aerial Images Trough Multispectral and LIDAR Data

The application of thematic maps obtained through the classification of remote images needs the obtained products with an optimal accuracy. The registered images from the airplanes display a very satisfactory spatial resolution, but the classical methods of thematic classification not always give better results than when the registered data from satellite are used. In order to improve these results of classification, in this work, the LIDAR sensor data from first return (Light Detection And Ranging) registered simultaneously with the spectral sensor data from airborne are jointly used. The final results of the thematic classification of the scene object of study have been obtained, quantified and discussed with and without LIDAR data, after applying different methods: Maximum Likehood Classification, Support Vector Machine with four different functions kernel and Isodata clustering algorithm (ML, SVM-L, SVM-P, SVM-RBF, SVM-S, Isodata). The best results are obtained for SVM with Sigmoide kernel. These allow the correlation with others different physical parameters with great interest like Manning hydraulic coefficient, for their incorporation in a GIS and their application in hydraulic modeling.

DAEDALUS at ImageCLEF 2011 Plant Identification Task: Using SIFT Keypoints for Object Detection

This paper describes the participation of DAEDALUS at ImageCLEF 2011 Plant Identification task. The task is evaluated as a supervised classification problem over 71 tree species from the French Mediterranean area used as class labels, based on visual content from scan, scan-like and natural photo images. Our approach to this task is to build a classifier based on the detection of keypoints from the images extracted using Lowe’s Scale Invariant Feature Transform (SIFT) algorithm. Although our overall classification score is very low as compared to other participant groups, the main conclusion that can be drawn is that SIFT keypoints seem to work significantly better for photos than for the other image types, so our approach may be a feasible strategy for the classification of this kind of visual content.

DAEDALUS at ImageCLEF Medical Retrieval 2011: Textual, Visual and Multimodal Experiments

This paper describes the participation of DAEDALUS at ImageCLEF 2011 Medical Retrieval task. We have focused on multimodal (or mixed) experiments that combine textual and visual retrieval. The main objective of our research has been to evaluate the effect on the medical retrieval process of the existence of an extended corpus that is annotated with the image type, associated to both the image itself and also to its textual description. For this purpose, an image classifier has been developed to tag each document with its class (1st level of the hierarchy: Radiology, Microscopy, Photograph, Graphic, Other) and subclass (2nd level: AN, CT, MR, etc.). For the textual-based experiments, several runs using different semantic expansion techniques have been performed. For the visual-based retrieval, different runs are defined by the corpus used in the retrieval process and the strategy for obtaining the class and/or subclass. The best results are achieved in runs that make use of the image subclass based on the classification of the sample images. Although different multimodal strategies have been submitted, none of them has shown to be able to provide results that are at least comparable to the ones achieved by the textual retrieval alone. We believe that we have been unable to find a metric for the assessment of the relevance of the results provided by the visual and textual processes

Spatial and thematic assessment of object- based forest stands delineation using OFA-matrix

Validación de la cartografía generada del terreno a partir de una nuevo sistema de validación propuesto

Novel parameters for evaluating the Spatial and Thematic accuracy of land cover maps.

Se proponen novedosas fórmulas para evaluar la certeza de la cartografía

DAEDALUS at SemEval-2014 Task 9: Comparing approaches for sentiment analysis in twitter

This paper describes our participation at SemEval- 2014 sentiment analysis task, in both contextual and message polarity classification. Our idea was to com- pare two different techniques for sentiment analysis. First, a machine learning classifier specifically built for the task using the provided training corpus. On the other hand, a lexicon-based approach using natural language processing techniques, developed for a ge- neric sentiment analysis task with no adaptation to the provided training corpus. Results, though far from the best runs, prove that the generic model is more robust as it achieves a more balanced evaluation for message polarity along the different test sets.

DAEDALUS at RepLab 2014: Detecting RepTrak reputation dimensions on tweets

This paper describes our participation at the RepLab 2014 reputation dimensions scenario. Our idea was to evaluate the best combination strategy of a machine learning classifier with a rule-based algorithm based on logical expressions of terms. Results show that our baseline experiment using just Naive Bayes Multinomial with a term vector model representation of the tweet text is ranked second among runs from all participants in terms of accuracy.

DAEDALUS at PAN 2014: Guessing tweet author's gender and age

This paper describes our participation at PAN 2014 author profiling task. Our idea was to define, develop and evaluate a simple machine learning classifier able to guess the gender and the age of a given user based on his/her texts, which could become part of the solution portfolio of the company. We were interested in finding not the best possible classifier that achieves the highest accuracy, but to find the optimum balance between performance and throughput using the most simple strategy and less dependent of external systems. Results show that our software using Naive Bayes Multinomial with a term vector model representation of the text is ranked quite well among the rest of participants in terms of accuracy.