1,720,971 research outputs found
Estimating the queue length at street intersections by using a movement feature space approach
Full text linkThis study aims to estimate the traffic load at street intersections obtaining the circulating vehicle number through image processing and pattern recognition. The algorithm detects moving objects in a street view by using level lines and generates a new feature space called movement feature space (MFS). The MFS generates primitives as segments and corners to match vehicle model generating hypotheses. The MFS is also grouped in a histogram configuration called histograms of oriented level lines (HO2 L). This work uses HO2 L features to validate vehicle hypotheses comparing the performance of different classifiers: linear support vector machine (SVM), non-linear SVM, neural networks and boosting. On average, successful detection rate is of 86% with 10-1 false positives per image for highly occluded images.Fil: Negri, Pablo Augusto. Universidad Argentina de la Empresa. Facultad de Ingeniería y Ciencias Exactas. Instituto de Tecnología; Argentina. Consejo Nacional de Investigaciones Científicas y Técnicas; Argentin
A MATLAB SMO implementation to train a SVM classifier: Application to multi-style license plate numbers recognition
Full text linkThis paper implements the Support Vector Machine (SVM) training procedure proposed by John Platt denominated Sequential Minimimal Optimization (SMO). The application of this system involves a multi-style license plate characters recognition identifying numbers from “0” to “9”. In order to be robust against license plates with different character/background colors, the characters (numbers) visual information is encoded using Histograms of Oriented Gradients (HOG). A reliability measure to validate the system outputs is also proposed. Several tests are performed to evaluate the sensitivity of the algorithm to different parameters and kernel functions.Fil: Negri, Pablo Augusto. Consejo Nacional de Investigaciones Científicas y Técnicas. Oficina de Coordinación Administrativa Ciudad Universitaria. Instituto de Investigación En Ciencias de la Computación. Universidad de Buenos Aires. Facultad de Ciencias Exactas y Naturales. Instituto de Investigación En Ciencias de la Computacion; Argentin
Detecting pedestrians on a Movement Feature Space
Full text linkThis work aims at detecting pedestrians in surveillance video sequences. A pre-processing step detects motion regions on the image using a scene background model based on level lines, which generates a Movement Feature Space, and a family of oriented histogram descriptors. A cascade of boosted classifiers generates pedestrian hypotheses using this feature space. Then, a linear Support Vector Machine validates the hypotheses that are likeliest to contain a person. The combination of the three detection phases reduces false positives, preserving the majority of pedestrians. The system tests conducted in our dataset, which contain low-resolution pedestrians, achieved a maximum performance of 25.5% miss rate with a rate of 10−1 false positives per image. This value is comparable to the best detection values for this kind of images. In addition, the processing time is between 2 and 6 fps on 640 480 pixel captures. This is therefore a fast and reliable pedestrian detector.Fil: Negri, Pablo Augusto. Universidad Arg.de la Empresa. Facultad de Ingeniería y Ciencias Exactas. Instituto de Tecnología; Argentina. Consejo Nacional de Investigaciones Científicas y Técnicas; ArgentinaFil: Goussies, Norberto Adrián. Universidad de Buenos Aires. Facultad de Ciencias Exactas y Naturales. Departamento de Computación; Argentina. Consejo Nacional de Investigaciones Científicas y Técnicas; ArgentinaFil: Lotito, Pablo Andres. Universidad Nacional del Centro de la Provincia de Buenos Aires. Facultad de Ciencias Exactas. Grupo de Plasmas Densos Magnetizados. Provincia de Buenos Aires. Gobernación. Comision de Investigaciones Científicas. Grupo de Plasmas Densos Magnetizados; Argentina. Consejo Nacional de Investigaciones Científicas y Técnicas; Argentin
Experimental characterization of collision avoidance in pedestrian dynamics
Full text linkIn the present paper, the avoidance behavior of pedestrians was characterized by controlled experiments. Several conflict situations were studied considering different flow rates and group sizes in crossing and head-on configurations. Pedestrians were recorded from above, and individual two-dimensional trajectories of their displacement were recovered after image processing. Lateral swaying amplitude and step lengths were measured for free pedestrians, obtaining similar values to the ones reported in the literature. Minimum avoidance distances were computed in two-pedestrian experiments. In the case of one pedestrian dodging an arrested one, the avoidance distance did not depend on the relative orientation of the still pedestrian with respect to the direction of motion of the first. When both pedestrians were moving, the avoidance distance in a perpendicular encounter was longer than the one obtained during a head-on approach. It was found that the mean curvature of the trajectories was linearly anticorrelated with the mean speed. Furthermore, two common avoidance maneuvers, stopping and steering, were defined from the analysis of the acceleration and curvature in single trajectories. Interestingly, it was more probable to observe steering events than stopping ones, also the probability of simultaneous steering and stopping occurrences was negligible. The results obtained in this paper can be used to validate and calibrate pedestrian dynamics models.Fil: Parisi, Daniel Ricardo. Instituto Tecnológico de Buenos Aires; Argentina. Consejo Nacional de Investigaciones Científicas y Técnicas; ArgentinaFil: Negri, Pablo Augusto. Universidad Argentina de la Empresa. Facultad de Ingeniería y Ciencias Exactas. Instituto de Tecnología; Argentina. Consejo Nacional de Investigaciones Científicas y Técnicas; ArgentinaFil: Bruno, Luciana. Consejo Nacional de Investigaciones Científicas y Técnicas. Oficina de Coordinación Administrativa Ciudad Universitaria. Instituto de Física de Buenos Aires. Universidad de Buenos Aires. Facultad de Ciencias Exactas y Naturales. Instituto de Física de Buenos Aires; Argentin
Toward making canopy hemispherical photography independent of illumination conditions: A deep-learning-based approach
No full textHemispherical photography produces the most accurate results when working with well-exposed photographs acquired under diffuse light conditions (diffuse-light images). Obtaining such data can be prohibitively expensive when surveying hundreds of plots is required. A relatively inexpensive alternative is using photographs acquired under direct sunlight (sunlight images). However, this practice leads to high errors since the standard processing algorithms expect diffuse-light imagery. Here, instead of using classification algorithms, which is the unique dominant practice, we approached the processing of sunlight images using deep learning (DL) regression. We implemented DL systems by using the general-purpose convolutional neural networks known as VGGNet 16, VGGNet 19, Res-Net, and SE-ResNet. We trained them with 608 samples acquired in a South American temperate forest populated by Nothofagus pumilio. For their evaluation, we used 113 independent samples. Each sample (X, Y) consisted of one or several sunlight images (X), and the plant area index (PAI) and effective PAI (PAIe) extracted from a diffuse-light image (Y). The sunlight images include clear sky and broken clouds with sun elevation from 15° to 47°. We obtained the best results with the SE-ResNet architecture. The system requires a low-resolution input reprojected to cylindrical, and it can make predictions with 10% root mean square error, even from pictures acquired with automatic exposure, which challenge previous findings. Furthermore, similar results (R2= 0.9, n = 104) can be obtained by feeding the system with photographs acquired with an inexpensive fisheye converter attached to a smartphone. Altogether, results indicate that our approach is a cost-efficient option for surveying hundreds of plots under direct sunlight. Therefore, combining our method with the traditional procedures provides processing solutions for virtually all kinds of illumination conditions.Fil: Díaz, Gastón Mauro. Centro de Investigación y Extensión Forestal Andino Patagónico; Argentina. Consejo Nacional de Investigaciones Científicas y Técnicas; ArgentinaFil: Negri, Pablo Augusto. Consejo Nacional de Investigaciones Científicas y Técnicas. Oficina de Coordinación Administrativa Ciudad Universitaria. Instituto de Investigación en Ciencias de la Computación; ArgentinaFil: Lencinas, José Daniel. Consejo Nacional de Investigaciones Científicas y Técnicas; Argentina. Centro de Investigación y Extensión Forestal Andino Patagónico; Argentin
Seguimiento de peatones utilizando campos probabilísticos y un espacio de descriptores dinámicos
Full text linkRecuperar información de secuencias de video, como la dinámica de peatones u otros objetos en movimiento en la escena, representa una herramienta indispensable para interpretar que está ocurriendo en la escena. Este artículo propone el uso de una Arquitectura basada en Targets, que asocian a cada persona una entidad autónoma y modeliza su dinámica con una máquina de estados. Nuestra metodología utiliza una familia de descriptores calculados en el Movement Feature Space (MFS) para realizar la detección y seguimiento de las personas. Esta arquitectura fue evaluada usando dos bases de datos públicas (PETS2009 y TownCentre), y comparándola con algoritmos de la literatura, arrojó mejores resultados, aun cuando estos algoritmos poseen una mayor complejidad computacional.Retrieving useful information from video sequences, such as the dynamics of pedestrians, and other moving objects on a video sequence, leads to further knowledge of what is happening on a scene. In this paper, a Target Framework associates each person with an autonomous entity, modeling its trajectory and speed by using a state machine. The particularity of our methodology is the use of a Movement Feature Space (MFS) to generate descriptors for classifiers and trackers. This approach is applied to two public sequences (PETS2009 and TownCentre). The results of this tracking outperform other algorithms reported in the literature, which have, however, a higher computational complexity.Fil: Negri, Pablo Augusto. Universidad Argentina de la Empresa. Facultad de Ingeniería y Ciencias Exactas. Instituto de Tecnología; Argentina. Consejo Nacional de Investigaciones Científicas y Técnicas; ArgentinaFil: Garayalde, Damian Alejandro. Consejo Nacional de Investigaciones Científicas y Técnicas; Argentina. Instituto Tecnológico de Buenos Aires; Argentin
Feature Fusion for Fingerprint Liveness Detection: a Comparative Study
Full text linkSpoofing attacks carried out using artificial replicas are a severe threat for fingerprint-based biometric systems and, thus, require the development of effective countermeasures. One possible protection method is to implement software modules that analyze fingerprint images to tell live from fake samples. Most of the static software-based approaches in the literature are based on various image features, each with its own strengths, weaknesses, and discriminative power. Such features can be seen as different and often complementary views of the object in analysis, and their fusion is likely to improve the classification accuracy. This paper aims at assessing the potential of these feature fusion approaches in the area of fingerprint liveness detection by analyzing different features and different methods for their aggregation. Experiments on publicly available benchmarks show the effectiveness of feature fusion methods, which improve the accuracy of those based on individual features and are competitive with respect to the alternative methods, such as the ones based on convolutional neural networks.Fil: Toosi, Amirhosein. Politecnico di Torino; ItaliaFil: Bottino, Andrea. Politecnico di Torino; ItaliaFil: Cumani, Sandro. Politecnico di Torino; ItaliaFil: Negri, Pablo Augusto. Consejo Nacional de Investigaciones Científicas y Técnicas; Argentina. Universidad de Buenos Aires; ArgentinaFil: Sottile, Pietro Luca. Politecnico di Torino; Itali
Vehicle make and model identification using vision system
Full text linkCet article présente un système de reconnaissance du type (constructeur, modèle) de véhicules par vision. À partir d’une vue de face avant d’un véhicule, limitée à sa calandre, nous en construisons une représentation à base de points de contour orientés. La classification est réalisée essentiellement en se fondant sur des algorithmes de votes. L’utilisation d’algorithmes de votes permet au système d’être robuste aux données manquantes ou erronées de la représentation. Nous avons donc construit une fonction de discrimination qui combine 3 votes et une distance, et agit comme une mesure de similarité entre chaque modèle et l’image de véhicule testée. Deux stratégies de décision ont été testées. La première associe à une image de calandre avant du véhicule, le modèle qui a obtenu la valeur la plus importante en sortie de la fonction. Une seconde stratégie regroupe toutes les sorties en un vecteur. La décision est alors prise via un algorithme de plus proche voisin dans un espace dit de votes. Avec la première stratégie, un taux de reconnaissance de 93 % est obtenu sur une base d’images prises en conditions réelles composée de 20 classes de type de véhicules. De plus, une caractérisation et une analyse du fonctionnement du système vis-à-vis de ses différents paramètres est proposée. Cependant ce taux chute à 80 % lorsque le nombre de modèles passe à 50 classes. Pour le même nombre de classes, la seconde stratégie permet d’obtenir un taux supérieur à 90 %.Many vision based Intelligent Transport Systems are dedicated to detect, track or recognize vehicles in image sequences. Three main applications can be distinguished. Firstly, embedded cameras allow to detect obstacles and to compute distances from the equiped vehicle. Secondly, road monitoring measures traffic flow, notifies the health services in case of an accident or informes the police in case of a driving fault. Finally, Vehicle based access control systems for buildings or outdoor sites have to authentify incoming (or outcoming) cars. Rather than these two systems, the third one uses often only the recognition of a small part of vehicle: the license plate. It is enough to identify a vehicle, but in practice the vision based number plate recognition system can provide a wrong information, due to a poor image quality or a fake plate. Combining such systems with others process dedicated to identify vehicle type (brand and model) the authentication can be increased in robustness. This paper adresses the identification problem of a vehicle type from a vehicle greyscale frontal image: the input of the system is an unknown vehicle class, that the system has to determine from a data base. This multiclass recognition system is developed using the oriented-contour pixels to represent each vehicle class. The system analyses a vehicle frontal view identifying the instance as the most similar model class in the data base. The classification is based on voting process and a Euclidean edge distance. The algorithm have to deal with partial occlusions. Tollgates hide a part of the vehicle and making inadequate the appearance-based methods. In spite of tollgate presence, our system doesn’t have to change the training base or apply time-consuming reconstruction process.Fil: Clady, Xavier. Université Pierre et Marie Curie-Paris 6. Institut des Systèmes Intelligents et de Robotique; FranciaFil: Negri, Pablo Augusto. Université Pierre et Marie Curie-Paris 6. Institut des Systèmes Intelligents et de Robotique; Francia. Consejo Nacional de Investigaciones Científicas y Técnicas. Oficina de Coordinación Administrativa Ciudad Universitaria. Instituto de Investigación en Ciencias de la Computación; ArgentinaFil: Milgram, Maurice. Université Pierre et Marie Curie-Paris 6. Institut des Systèmes Intelligents et de Robotique; FranciaFil: Poulenard, Raphael. LPREditor; Franci
Going Beyond Counting First Authors in Author Co-citation Analysis
Full text linkThe present study examines one of the fundamental aspects of author co-citation analysis (ACA) - the way co-citation
counts are defined. Co-citation counting provides the data on which all subsequent statistical analyses and mappings
are based, and we compare ACA results based on two different types of co-citation counting - the traditional type that
only counts the first one among a cited work's authors on the one hand and a non-traditional type that takes into
account the first 5 authors of a cited work on the other hand. Results indicate that the picture produced through this non-traditional author co-citation counting contains more coherent author groups and is therefore considerably clearer. However, this picture represents fewer specialties in the research field being studied than that produced through the traditional first-author co-citation counting when the same number of top-ranked authors is selected and analyzed. Reasons for these effects are discussed
- …
