Search

The human visual system (HVS) can process large quantities of visual information instantly. Visual saliency perception is the process of locating and identifying regions with a high degree of saliency from a visual standpoint. Mesh saliency detection has been studied extensively in recent years, but few studies have focused on 3D point cloud saliency detection. The estimation of visual saliency is important for computer graphics tasks such as simplification, segmentation, shape matching and resizing. In this paper, we present a method for the direct detection of saliency on unorganized point clouds. First, our method computes a set of overlapping neighborhoods and estimates a descriptor vector for each point inside it. Then, the descriptor vectors are used as a natural dictionary in order to apply a sparse coding process. Finally, we estimate a saliency map of the point neighborhoods based on the Minimum Description Length (MDL) principle. Experiment results show that the proposed method achieves similar results to those from the literature review and in some cases even improves on them. It captures the geometry of the point clouds without using any topological information and achieves an acceptable performance. The effectiveness and robustness of our approach are shown by comparing it to previous studies in the literature review.

En este paper se presenta un procedimiento rápido para la selección automática de centros de interpolación de funciones de base radial. El procedimiento se basa en la clusterización de la superficie del objeto en regiones con características geométricas homogéneas tales como: la estimación de curvatura. Este enfoque de clusterización garantiza que la totalidad de las variaciones de la superficie serán representadas en la construcción del interpolante. Posteriormente, un conjunto de reglas es aplicado a cada uno de los cluster, para seleccionar finalmente los puntos del cluster que deberán ser utilizados como centros de interpolación. Las reglas fueron extraídas de soluciones óptimas obtenidas mediante una Estrategia Evolutiva. Excelentes resultados sobre datos de rango reales son mostrados.

Land cover classification, generated from satellite imagery through semantic segmentation, has become fundamental for monitoring land use and land cover change (LULCC). The tropical Andes territory provides opportunities due to its significance in the provision of ecosystem services. However, the lack of reliable data for this region, coupled with challenges arising from its mountainous topography and diverse ecosystems, hinders the description of its coverage. Therefore, this research proposes the Tropical Andes Land Cover Dataset (TALANDCOVER). It is constructed from three sample strategies: aleatory, minimum 50%, and 70% of representation per class, which address imbalanced geographic data. Additionally, the U-Net deep learning model is applied for enhanced and tailored classification of land covers. Using high-resolution data from the NICFI program, our analysis focuses on the Department of Antioquia in Colombia. The TALANDCOVER dataset, presented in TIF format, comprises multiband R-G-B-NIR images paired with six labels (dense forest, grasslands, heterogeneous agricultural areas, bodies of water, built-up areas, and bare-degraded lands) with an estimated 0.76 F1 score compared to ground truth data by expert knowledge and surpassing the precision of existing global cover maps for the study area. To the best of our knowledge, this work is a pioneer in its release of open-source data for segmenting coverages with pixel-wise labeled NICFI imagery at a 4.77 m resolution. The experiments carried out with the application of the sample strategies and models show F1 score values of 0.70, 0.72, and 0.74 for aleatory, balanced 50%, and balanced 70%, respectively, over the expert segmented sample (ground truth), which suggests that the personalized application of our deep learning model, together with the TALANDCOVER dataset offers different possibilities that facilitate the training of deep architectures for the classification of large-scale covers in complex areas, such as the tropical Andes. This advance has significant potential for decision making, emphasizing sustainable land use and the conservation of natural resources.

Vegetation encroachment in power line corridors remains a major challenge for modern energy-dependent societies, as it can cause power outages and lead to significant financial losses. Unmanned Aerial Vehicles (UAVs) have emerged as a promising solution for monitoring infrastructure, owing to their ability to acquire high-resolution overhead images of these areas quickly and affordably. However, accurate segmentation of the vegetation encroachment in this imagery is a challenging task, due to the complexity of the scene and the high pixel imbalance between the power lines, the vegetation and the background classes. In this paper, we propose a deep learning-based approach to tackle this problem caused by the original and different geometry of the objects. Specifically, we use DeepLabV3, U-Net and a modified version of the U-Net architecture with VGG-16 weights to train two separate models. One of them segments the dominant classes, the vegetation from the background, achieving an IoU of 0.77. The other one segments power line corridors from the background, obtaining an IoU of 0.64. Finally, ensembling both models into one creates an “encroachment” zone, where power lines and vegetation are intersected. We train our models using the Vegetation Encroachment in Power Line Corridors dataset (VEPL), which includes RGB orthomosaics and multi-label masks for segmentation. Experimental results demonstrate that our approach outperforms individual networks and original prominent architectures when applied to this specific problem. This approach has the potential to significantly improve the efficiency and accuracy of vegetation encroachment monitoring using UAV, thus helping to ensure the reliability and sustainability of power supply.

The COVID-19 pandemic has had a significant impact on various aspects of society, including economic, health, political, and work-related domains. The pandemic has also caused an emotional effect on individuals, reflected in their opinions and comments on social media platforms, such as Twitter. This study explores the evolution of sentiment in Spanish pandemic tweets through a data analysis based on a fine-tuned BERT architecture. A total of six million tweets were collected using web scraping techniques, and pre-processing was applied to filter and clean the data. The fine-tuned BERT architecture was utilized to perform sentiment analysis, which allowed for a deep-learning approach to sentiment classification. The analysis results were graphically represented based on search criteria, such as “COVID-19” and “coronavirus”. This study reveals sentiment trends, significant concerns, relationship with announced news, public reactions, and information dissemination, among other aspects. These findings provide insight into the emotional impact of the COVID-19 pandemic on individuals and the corresponding impact on social media platforms.

Drone imagery is becoming the main source of overhead information to support decisions in many different fields, especially with deep learning integration. Datasets to train object detection and semantic segmentation models to solve geospatial data analysis are called GeoAI datasets. They are composed of images and corresponding labels represented by full-size masks typically obtained by manual digitizing. GIS software is made of a set of tools that can be used to automate tasks using geo-referenced raster and vector layers. This work describes a workflow using GIS tools to produce GeoAI datasets. In particular, it mentions the steps to obtain ground truth data from OSM and use methods for geometric and spectral augmentation and the data fusion of drone imagery. A method semi-automatically produces masks for point and line objects, calculating an optimum buffer distance. Tessellation into chips, pairing and imbalance checking is performed over the image–mask pairs. Dataset splitting into train–validation–test data is done randomly. All of the code for the different methods are provided in the paper, as well as point and road datasets produced as examples of point and line geometries, and the original drone orthomosaic images produced during the research. Semantic segmentation results performed over the point and line datasets using a classical U-Net show that the semi-automatically produced masks, called primitive masks, obtained a higher mIoU compared to other equal-size masks, and almost the same mIoU metric compared to full-size manual masks.

Detection and Semantic Segmentation of vehicles in drone aerial orthomosaics has applications in a variety of fields such as security, traffic and parking management, urban planning, logistics, and transportation, among many others. This paper presents the HAGDAVS dataset fusing RGB spectral channel and Digital Surface Model DSM for the detection and segmentation of vehicles from aerial drone images, including three vehicle classes: cars, motorcycles, and ghosts (motorcycle or car). We supply DSM as an additional variable to be included in deep learning and computer vision models to increase its accuracy. RGB orthomosaic, RG-DSM fusion, and multi-label mask are provided in Tag Image File Format. Geo-located vehicle bounding boxes are provided in GeoJSON vector format. We also describes the acquisition of drone data, the derived products, and the workflow to produce the dataset. Researchers would benefit from using the proposed dataset to improve results in the case of vehicle occlusion, geo-location, and the need for cleaning ghost vehicles. As far as we know, this is the first openly available dataset for vehicle detection and segmentation, comprising RG-DSM drone data fusion and different color masks for motorcycles, cars, and ghosts.

In recent years, the use of deep learning-based models for developing advanced healthcare systems has been growing due to the results they can achieve. However, the majority of the proposed deep learning-models largely use convolutional and pooling operations, causing a loss in valuable data and focusing on local information. In this paper, we propose a deep learning-based approach that uses global and local features which are of importance in the medical image segmentation process. In order to train the architecture, we used extracted three-dimensional (3D) blocks from the full magnetic resonance image resolution, which were sent through a set of successive convolutional neural network (CNN) layers free of pooling operations to extract local information. Later, we sent the resulting feature maps to successive layers of self-attention modules to obtain the global context, whose output was later dispatched to the decoder pipeline composed mostly of upsampling layers. The model was trained using the Mindboggle-101 dataset. The experimental results showed that the self-attention modules allow segmentation with a higher Mean Dice Score of 0.90 ± 0.036 compared with other UNet-based approaches. The average segmentation time was approximately 0.038 s per brain structure. The proposed model allows tackling the brain structure segmentation task properly. Exploiting the global context that the self-attention modules incorporate allows for more precise and faster segmentation. We segmented 37 brain structures and, to the best of our knowledge, it is the largest number of structures under a 3D approach using attention mechanisms.

According to worldwide statistics, traffic accidents are the cause of a high percentage of violent deaths. The time taken to send the medical response to the accident site is largely affected by the human factor and correlates with survival probability. Due to this and the wide use of video surveillance and intelligent traffic systems, an automated traffic accident detection approach becomes desirable for computer vision researchers. Nowadays, Deep Learning (DL)-based approaches have shown high performance in computer vision tasks that involve a complex features relationship. Therefore, this work develops an automated DL-based method capable of detecting traffic accidents on video. The proposed method assumes that traffic accident events are described by visual features occurring through a temporal way. Therefore, a visual features extraction phase, followed by a temporary pattern identification, compose the model architecture. The visual and temporal features are learned in the training phase through convolution and recurrent layers using built-from-scratch and public datasets. An accuracy of 98% is achieved in the detection of accidents in public traffic accident datasets, showing a high capacity in detection independent of the road structure.

High-resolution 3D scanning devices produce high-density point clouds, which require a large capacity of storage and time-consuming processing algorithms. In order to reduce both needs, it is common to apply surface simplification algorithms as a preprocessing stage. The goal of point cloud simplification algorithms is to reduce the volume of data while preserving the most relevant features of the original point cloud. In this paper, we present a new point cloud feature-preserving simplification algorithm. We use a global approach to detect saliencies on a given point cloud. Our method estimates a feature vector for each point in the cloud. The components of the feature vector are the normal vector coordinates, the point coordinates, and the surface curvature at each point. Feature vectors are used as basis signals to carry out a dictionary learning process, producing a trained dictionary. We perform the corresponding sparse coding process to produce a sparse matrix. To detect the saliencies, the proposed method uses two measures, the first of which takes into account the quantity of nonzero elements in each column vector of the sparse matrix and the second the reconstruction error of each signal. These measures are then combined to produce the final saliency value for each point in the cloud. Next, we proceed with the simplification of the point cloud, guided by the detected saliency and using the saliency values of each point as a dynamic clusterization radius. We validate the proposed method by comparing it with a set of state-of-the-art methods, demonstrating the effectiveness of the simplification method.

Denoising the point cloud is fundamental for reconstructing high quality surfaces with details in order to eliminate noise and outliers in the 3D scanning process. The challenges for a denoising algorithm are noise reduction and sharp features preservation. In this paper, we present a new model to reconstruct and smooth point clouds that combine L1-median filtering with sparse L1 regularization for both denoising the normal vectors and updating the position of the points to preserve sharp features in the point cloud. The L1-median filter is robust to outliers and noise compared to the mean. The L1 norm is a way to measure the sparsity of a solution, and applying an L1 optimization to the point cloud can measure the sparsity of sharp features, producing clean point set surfaces with sharp features. We optimize the L1 minimization problem by using the proximal gradient descent algorithm. Experimental results show that our approach is comparable to the state-of-the-art methods, as it filters out 3D models with a high level of noise, but keeps their geometric features.

Advances in data acquisition technology have made it possible to obtain high-density samples from the surface of an object. Such samples produce thousands or even millions of data points. Processing such large amounts of data is computationally expensive. This study presents a novel method for point cloud simplification using an estimated local density of the point cloud. The proposed approach is robust to noise and outliers. The method is comprised of three stages. The first stage uses the expectation maximization algorithm to cluster the point cloud according to the local distribution of the points. The second stage identifies the points with a high curvature. These are feature points that will not be removed. In the final stage, a linear programming model is applied to reduce the cloud. Each cluster is a graph where the nodes have a cost defined by the inverse of its distance to the centroid. The results show that the reduced cloud is a good approximation of the original. [ABSTRACT FROM AUTHOR]

Resumen En este artículo, se presenta una estrategia para el desarrollo de observatorios de riesgo de desastres, integrando inteligencia empresarial y análisis visual. El análisis y la contextualización a nivel global se llevan a cabo desde la perspectiva de políticas, herramientas y literatura científica en el contexto de la reducción del riesgo de desastres. La estrategia integra la metodología Kimball para la construcción de proyectos de inteligencia empresarial (BI) y el marco de visualización de Tamara Munzner para fortalecer el proceso de toma de decisiones y hacerlo flexible y más rápido, satisfaciendo las necesidades de los usuarios empresariales. Considerar la gestión del riesgo de desastres como un proceso permite la identificación del punto de partida y los procesos que siguen, lo que resulta en un inicio rápido del ciclo de vida del proyecto. Además, incluir el marco de visualización, junto con la metodología de BI, facilita el desarrollo de herramientas de análisis para resolver los problemas del usuario final.

Model-Driven Development (MDD) is a software development approach that facilitates problem comprehension. MDD is carried out based on level of abstraction attained by working with models. When using models to create user-interfaces, development time can be reduced significantly. However, automatic user-interface generation is done with preestablished templates that might not fulfill all the requirements clients. These templates might also be too general and sometimes provide few customization options. In this paper, we review research on the usage of MDD for user-interface development. We also study how automatic user-interface generation can result in limitations as the need might arise for repeatedly modifying model and code.

Multiple rainfall-landslides relationships have been proposed for early warning systems in the world. This work presents a data base that synthesizes the state of the art of rainfall thresholds that generates mass movements in the entire world. The more common rainfall-landslides relationships are processed in order to analyze data in function of the Köppen climatic classification and the area for which it was proposed. The results show a wide range of parameters derived from rainfall that triggers landslides together with considerable differences between threshold values. Data analysis shows the necessity to obtain parameters and thresholds derived from the rainfall that generate landslides for each specific geographical and climatic zone. For that reason, it is possible to generate a rational background that allows us to establish early warning systems of landslides caused by rainfall with low variability. This work provides insights into the catalog of possible parameters derived from rainfall records that allow us to generate thresholds to predict the landslides occurrence. Researchers will have the possibility to search for rainfall-landslide relationships based on statistical analysis, with a wide spectrum of parameters, beyond the traditional variables of antecedent rainfall and the intensity occurred the day El modelado y simulación de pequeños sistemas industriales ha adquirido gran importancia en los procesos de toma de decisiones, toda vez que les permite a microempresas fundamentar sus decisiones de producción. Este artículo hace un análisis de una empresa dedicada a la confección de ropa exterior masculina con una producción semanal promedio de 490 prendas. La motivación del estudio radica en el interés de incrementar su productividad; por esto, se desarrolló un modelo del sistema real con la herramienta Arena®, logrando identificar las partes de los procesos responsables de atrasos y “cuellos de botella” que permitieron plantear el escenario de solución. Este documento muestra los resultados de la simulación del sistema real y una comparación con un escenario de mejora que presentó un aumento en la productividad. Palabras clave: Simulación Discreta, Modelamiento, Confecciones, Toma de Decisiones, Arena®.f the landslide.

Este documento reporta los resultados de una investigación orientada hacia el diseño de un método de tratamiento de imágenes digitales para la automatización de los procesos de registro y control de marcas de ganado requeridas por las regulaciones del sector ganadero en Colombia. El método permite automatizar los procesos de búsqueda y de comparación necesarios para garantizar la unicidad de las marcas dentro de un sistema asistido por computadora. Se inicia con la generación de un histograma estimado de la geometría de la marca, lo que permite comparar y detectar similitudes entre las figuras previamente almacenadas, mediante una métrica de similitud basada en la distancia de Minkowski. Los resultados obtenidos indican que el método es adecuado para realizar un proceso de discriminación de dichas imágenes, reducir las ambigüedades y garantizar la unicidad de los registros. Los resultados obtenidos y un análisis de su aplicación son reportados.

En este artículo se desarrolla un nuevo esquema de cuatro puntos para la subdivisión interpolante de curvas basado en la primera derivada discreta (DFDS), el cual, reduce la formación de oscilaciones indeseables que pueden surgir en la curva límite cuando los puntos de control no obedecen a una parametrización uniforme. Se empleó un conjunto de 3000 curvas cuyos puntos de control fueron generados aleatoriamente. Curvas suaves fueron obtenidas tras siete pasos de subdivisión empleando los esquemas DFDS, Cuatro-puntos (4P), Nuevo de cuatro-puntos (N4P), Cuatro-puntos ajustado (T4P) y el Esquema interpolante geométricamente controlado (GC4P). Sobre cada curva suave se evaluó la propiedad de tortuosidad. Un análisis de las distribuciones de frecuencia obtenidas para esta propiedad, empleando la prueba de KruskalWallis, revela que el esquema DFDS posee los menores valores de tortuosidad en un rango más estrecho.

En Colombia la industria de la exportación de frutas tropicales, en particular la del mango de azúcar, se caracteriza por el tratamiento manual en cada una de las etapas previas a su comercialización, particularmente, la aplicación de los procedimientos de clasificación acordes con los criterios estandarizados nacionales de calidad (NTC-5139). En este artículo de investigación se presenta el diseño de un prototipo mecánico para automatizar el proceso de clasificación del mango de azúcar (Mangifera Indica L) tipo exportación. Este prototipo está constituido por un sistema de bandas que permite transportar y rotar el fruto para que un sistema de adquisición capture diferente caras de este. Las imágenes obtenidas se analizaron mediante técnicas computacionales que permiten obtener una aproximación tridimensional, lo que posibilita la estimación del volumen de cada fruto. Los resultados de la operación del mecanismo, y la aplicación de los métodos mostraron una precisión superior al 90%.In Colombia, tropical fruit export industry, specifically ''mango de azúcar'' (sweet mango), is characterized for the manual treatment in each stage prior to commercialization, specially application of classification procedures according to local quality standardized criteria (NTC-5139). This research article describes the design of a mechanical prototype for automating the export-type ''mango de azúcar'' (Mangifera Indica L) classification process. This prototype consists of a band system which allows conveying and rotating the fruit so that an acquisition system captures several faces of the fruit. Images obtained were analyzed through computational techniques which allow obtaining a 3-D approach, and this makes estimation of each fruit volume possible. Results of the operation of this mechanisms and the application of methods showed an accuracy over 90%.

En este trabajo se propone un método de registro grueso de mallas triangulares para la reconstrucción de objetos tridimensionales, empleando el índice de forma para obtener información topológica de la superficie y la comparación de regiones con forma similar empleando la métrica Earth Mover's Distance (EMD). Las principales ventajas de este método es que permite independencia de la pre-alineación inicial y transformación espacial, permitiendo obtener un registro adecuado sin importar la posición inicial de cada vista.

Es ampliamente conocido que algunos campos relacionados con aplicaciones de gráficos realistas requieren modelos tridimensionales altamente detallados. Las tecnologías para esto están bien desarrolladas, sin embargo, en algunos casos los escáneres láser obtienen modelos complejos formados por millones de puntos, por lo que son computacionalmente intratables. En estos casos es conveniente obtener un conjunto reducido de estas muestras con las que reconstruir la superficie de la función. Obtener un enfoque de reducción adecuado que posea un equilibrio entre la pérdida de precisión de la función reconstruida, y el costo computacional es un problema no trivial. En este artículo presentamos un método jerárquico de aglomeración a través de la selección de centros mediante la geométrica, la distribución y la estimación de curvatura de las muestras en el espacio 3D.It is widely known that some fields related to graphic applications require realistic and full detailed three-dimensional models. Technologies for this kind of applications exist. However, in some cases, laser scanner get complex models composed of million of points, making its computationally difficult. In these cases, it is desirable to obtain a reduced set of these samples to reconstruct the function's surface. An appropriate reduction approach with a non-significant loss of accuracy in the reconstructed function with a good balance of computational load is usually a non-trivial problem. In this article, a hierarchical clustering based method by the selection of center using the geometric distribution and curvature estimation of the samples in the 3D space is described.

Los diferentes métodos de ajuste de superficies desde datos dispersos asumen que la superficie reconstruida no presenta ningún tipo de discontinuidad. Las discontinuidades en las superficies son obtenidas a partir de la forma del objeto original y estas aparecen principalmente en el proceso de adquisición debido a problemas de oclusión. Tradicionalmente, para la representación de superficies que tienen discontinuidades se realiza un proceso manual adicional después de la etapa final de ajuste. En este artículo se propone un método para detectar y preservar discontinuidades automáticamente en superficies de topología arbitraria, así como un método para la reducción de los puntos a interpolar, que permite disminuir el costo computacional asociado a la solución del interpolante. Inicialmente, el método realiza una clasificación basada en la estimación de la varianza de la torsión de la curva del contorno. Cuando el valor de la varianza en el contorno de los huecos es muy bajo, entonces se considera que este hueco debe ser preservado. Luego, la superficie descrita por el conjunto de puntos seleccionados es interpolada usando Funciones de Base Radial. Por último, la superficie generada es cortada con cada contorno identificado como preservable.Different methods for fitting surfaces from dispersed data interpolation assume the reconstructed surface does not have any type of discontinuity. Discontinuities in the surfaces are obtained from the original shape of the object and they appear mainly in the acquisition process due to occlusion problems. For digital representation of surfaces with discontinuities is commonly found an additional manual process, which is applied after the final stage of fitting surface. In this paper, two methods are proposed, one of them with the aim of detect and preserve discontinuities in surfaces with arbitrary topology, and the other one to reduce the amount of points to be interpolated, decreasing the computational cost. Initially, the method carries out a classification based on estimation of the torsion variance of the contour curve. When the holes contour has very low values of variances, it is considered that this hole must be preserved. Then, the surface described by the selected set of points is interpolated using Radial Basis Functions. Finally, the generated surface is trimmed with each contour identified like preservable.

Recientemente en la comunidad de visión por computador relacionada con los procesos de reconstrucción de superficies de objetos de forma libre se ha notado una tendencia creciente en el uso de técnicas de interpolación. En esta área los interpoladores de funciones de base radial han mostrado que producen modelos tridimensionales con altos niveles de precisión, gran flexibilidad en la reproducción de formas complejas y una alta tolerancia a los niveles de ruido. Las funciones de base radial necesitan la estimación de un conjunto de parámetros de los cuales depende la correcta reconstrucción de la superficie. En este trabajo se describe un procedimiento para la reconstrucción de superficies mediante interpolación usando funciones de base radial de soporte compacto realizando una estimación de centros y radios basado en las características geométricas de la superficie a reconstruir. Los resultados del método propuesto son descritos mediante la interpolacion de datos de rango de objetos reales.

En este artículo se presenta un procedimiento para la detección del disco óptico (DO) en retinografías, mediante un algoritmo evolutivo. El procedimiento tiene dos etapas principales: la detección gruesa de la posición del DO y el refinamiento de los bordes del contorno. La detección gruesa ubica la posición del DO mediante un algoritmo evolutivo, cuyos individuos tienen como función objetivo la cantidad de píxeles brillantes y el número de bordes de la red de conductos sanguíneos, contenidos dentro de una circunferencia. La etapa de refinamiento aplica un procedimiento geométrico, para deformar el círculo inicial, ajustando el borde de éste con la posición del píxel de mayor variación en dirección al vector normal. El procedimiento fue evaluado empleando los repositorios públicos STARE y DIAREDB, procesando imágenes de pacientes sanos y con alteraciones de las retina, generadas por la presencia de retinopatía diabética. Los resultados experimentales muestran que el método propuesto puede identificar la posición del disco óptico en retinografías con una precisión cercana al 96 %.