Nallig Leal - Academia.edu (original) (raw)
Papers by Nallig Leal
Congreso Internacional de Inteligencia Computacional, 2005
Image and Vision Computing, 2021
Recently, intelligent video surveillance applications have become essential in public security by... more Recently, intelligent video surveillance applications have become essential in public security by the use of computer vision technologies to investigate and understand long video streams. Anomaly detection and classification are considered a major element of intelligent video surveillance. The aim of anomaly detection is to automatically determine the existence of abnormalities in a short time period. Deep reinforcement learning (DRL) techniques can be employed for anomaly detection, which integrates the concepts of reinforcement learning and deep learning enabling the artificial agents in learning the knowledge and experience from actual data directly. With this motivation, this paper presents an Intelligent Video Anomaly Detection and Classification using Faster RCNN with Deep Reinforcement Learning Model, called IVADC-FDRL model. The presented IVADC-FDRL model operates on two major stages namely anomaly detection and classification. Firstly, Faster RCNN model is applied as an object detector with Residual Network as a baseline model, which detects the anomalies as objects. Besides, deep Q-learning (DQL) based DRL model is employed for the classification of detected anomalies. In order to validate the effective anomaly detection and classification performance of the IVADC-FDRL model, an extensive set of experimentations were carried out on the benchmark UCSD anomaly dataset. The experimental results showcased the better performance of the IVADC-FDRL model over the other compared methods with the maximum accuracy of 98.50% and 94.80% on the applied Test004 and Test007 dataset respectively.
Automatic recognition of an acoustic signature in underwater environments is an important and act... more Automatic recognition of an acoustic signature in underwater environments is an important and active field with multiple applications, one of which is vessel recognition. When a vessel moves through the sea, its engine and the cavitation generated by its propellers produce an acoustic wave of unique characteristics that allow for its individual identification. The problem of identification involves several variables, such as ambient noise, biological noise, and even noise produced by its own machinery, which means that the signal produced, is complex to treat. This paper presents a method based on Fourier transform and digital signal processing to extract a set of features allowing for automatic ship classification (by type). Computational intelligence techniques such as Artificial Neural Networks (ANNs) and Support Vector Machines (SVMs) are used for the classification stage. Results showed that the vessel recognition system has accuracy close to 92%.
Este trabajo esta relacionado con las anomalias representadas por discontinuidades o huecos prese... more Este trabajo esta relacionado con las anomalias representadas por discontinuidades o huecos presentes en nubes de puntos tridimensionales. Las auto-oclusiones y las propiedades opticas del material del objeto sensado constituyen las fuentes principales de anomalias. El presente estudio fue realizado con el proposito de analizar similitudes o disimilitudes estadisticas de las propiedades geometricas de los contornos generados por estas dos fuentes. Las propiedades analizadas fueron la estimacion de la curvatura y la torsion, para esto se clasificaron los diferentes contornos en dos grupos: los generados por propiedades opticas del material del objeto y los generados por oclusion. Los resultados sugieren que estas fuentes generan contornos de discontinuidades estadisticamente similares, dificultando una discriminacion basada en estas propiedades.
Rev. Avances en Sistemas Informática, 2010
Traffic surveillance systems promise to improve traffic safety and alleviate traffic jams, among ... more Traffic surveillance systems promise to improve traffic safety and alleviate traffic jams, among other things. The continuous increase in processing speed of computers in the last decade has led to an increasing emphasis on research and development of automatic traffic surveillance systems in different countries. However, traffic surveillance systems are still an open research area because they are not in a state of maturity that allows them to operate with complete autonomy in the task of regulating traffic. This article presents a review of automatic traffic surveillance systems based on vehicular image segmentation techniques. We present a classification of these systems according to their purpose and nature, and a classification of the most commonly used techniques in such systems.
Sensors (Basel, Switzerland), 2021
High-resolution 3D scanning devices produce high-density point clouds, which require a large capa... more 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 s...
Rev. Avances en Sistemas Informática, 2009
Este articulo presenta un nuevo metodo de simplificacion de nubes de puntos. El metodo propuesto,... more Este articulo presenta un nuevo metodo de simplificacion de nubes de puntos. El metodo propuesto, a diferencia de otros, no requiere la construccion previa de mallas poligonales y es robusto al ruido y a valores atipicos presentes en los datos. El metodo propuesto se compone principalmente de tres etapas. En la primera etapa, se segmenta la nube de puntos en regiones homogeneas, usando el algoritmo kmeans. En la segunda etapa , se ajusta un plano de regresion de componentes principales robusto al ruido en cada cluster para determinar la tendencia local de los puntos. Finalmente, en la tercera etapa, usando un algoritmo genetico se seleccionan los puntos de cada cluster cuyo plano de regresion de analisis de componentes principales minimice el angulo con el plano de regresion del cluster. Resultados exper imentales muestran que la distribucion local y global de la nube de puntos original se mantiene.
2019 14th Iberian Conference on Information Systems and Technologies (CISTI), 2019
The human brain and the human vision system, in front of a given scene, focuses on regions with m... more The human brain and the human vision system, in front of a given scene, focuses on regions with more information. Many pieces of research in the field of psychology, neuropsychology and cognitive neurosciences, have detailed the way in which the extraction of such information is carried out, even proposing models of the functioning of visual attention. From these models, computer scientists have proposed computational variants, which imitate human visual attention. This field of research is known as saliency detection on images. In this paper, a new and simple approach for detecting visual saliencies based on Dictionary Learning and Sparse Coding is presented. Our method first subdivides the image into full overlapped patches and runs a dictionary learning over them for obtaining its sparse representation. Then, by analyzing the sparse coding matrix, we compute how many image patches a dictionary atom affects in order to classify them as frequent or rare. Then, we calculate the sali...
Different factors affect the objects that are considered cultural heritage of the regions, among ... more Different factors affect the objects that are considered cultural heritage of the regions, among the most common causes we can find the environmental conditions, long exposure to light, fungi presence, among others. The cultural heritage objects conservation is desirable from the regional cultural importance point of view, therefore, is interesting to explore new technological-based techniques addressed to the conservation. This paper presents an application case of archeological pieces three-dimensional reconstruction using scanner based on structuredlight. It describes the whole procedure including the stages of data acquisition, preprocessing, registration, anomalies correction and visualization. These stages were making using open software, thus providing a precise and cost effective solution. The described setup allows to reproduce digitally an accurate three-dimensional model without need specialized software and hardware.
Automatic recognition of an acoustic signature in underwater environments is an important and act... more Automatic recognition of an acoustic signature in underwater environments is an important and active field with multiple applications, one of which is vessel recognition. When a vessel moves through the sea, its engine and the cavitation generated by its propellers produce an acoustic wave of unique characteristics that allow for its individual identification. The problem of identification involves several variables, such as ambient noise, biological noise, and even noise produced by its own machinery, which means that the signal produced, is complex to treat. This paper presents a method based on Fourier transform and digital signal processing to extract a set of features allowing for automatic ship classification (by type). Computational intelligence techniques such as Artificial Neural Networks (ANNs) and Support Vector Machines (SVMs) are used for the classification stage. Results showed that the vessel recognition system has accuracy close to 92%.
Rev. Avances en Sistemas Informática, 2009
In this paper is presented a method to surface segmentation from sampled points, without a previo... more In this paper is presented a method to surface segmentation from sampled points, without a previous surface approximation, like triangular mesh or local polynomial regression. The segmentation process proceed from the information give by the data points, by this reason are necessary use algorithms, that estimates geometric properties like normals, curvatures and detection of surface discontinuities without its approximations. This peper presents a method to estimate the normals to each point in the surface, taking in account the local
Recently, intelligent video surveillance applications have become essential in public security by... more Recently, intelligent video surveillance applications have become essential in public security by the use of computer vision technologies to investigate and understand long video streams. Anomaly detection and classification are considered a major element of intelligent video surveillance. The aim of anomaly detection is to automatically determine the existence of abnormalities in a short time period. Deep reinforcement learning (DRL) techniques can be employed for anomaly detection, which integrates the concepts of reinforcement learning and deep learning enabling the artificial agents in learning the knowledge and experience from actual data directly. With this motivation, this paper presents an Intelligent Video Anomaly Detection and Classification using Faster RCNN with Deep Reinforcement Learning Model, called IVADC-FDRL model. The presented IVADC-FDRL model operates on two major stages namely anomaly detection and classification. Firstly, Faster RCNN model is applied as an obj...
Advances in data acquisition technology have made it possible to obtain high-density samples from... more 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.
Sensors
Magnetic Resonance (MR) Imaging is a diagnostic technique that produces noisy images, which must ... more Magnetic Resonance (MR) Imaging is a diagnostic technique that produces noisy images, which must be filtered before processing to prevent diagnostic errors. However, filtering the noise while keeping fine details is a difficult task. This paper presents a method, based on sparse representations and singular value decomposition (SVD), for non-locally denoising MR images. The proposed method prevents blurring, artifacts, and residual noise. Our method is composed of three stages. The first stage divides the image into sub-volumes, to obtain its sparse representation, by using the KSVD algorithm. Then, the global influence of the dictionary atoms is computed to upgrade the dictionary and obtain a better reconstruction of the sub-volumes. In the second stage, based on the sparse representation, the noise-free sub-volume is estimated using a non-local approach and SVD. The noise-free voxel is reconstructed by aggregating the overlapped voxels according to the rarity of the sub-volumes it...
Ingenieria y Universidad
Introduction: The use of mobile applications has increased in the last years. Most of them requir... more Introduction: The use of mobile applications has increased in the last years. Most of them require the knowledge of the user location, either for their core service or for marketing purposes. Location-based services (LBS) offer context-based assistance to users based on their location. Although these applications ask the user for permission to use their location and even explain in detail how this information will be used in its terms and conditions, most users are not aware or even interested in the fact that their location information is stored in databases and monetized by selling it to third-party companies. Regarding this situation, we developed a study with the aim to assess perception, concerns and awareness from users about their location information. Methods: This work is based on an exploratory survey applied to the university community, mainly from the North Coast of Colombia, to measure the perception of location privacy of users with mobile devices. The questionnaire wa...
Congreso Internacional de Inteligencia Computacional, 2005
Image and Vision Computing, 2021
Recently, intelligent video surveillance applications have become essential in public security by... more Recently, intelligent video surveillance applications have become essential in public security by the use of computer vision technologies to investigate and understand long video streams. Anomaly detection and classification are considered a major element of intelligent video surveillance. The aim of anomaly detection is to automatically determine the existence of abnormalities in a short time period. Deep reinforcement learning (DRL) techniques can be employed for anomaly detection, which integrates the concepts of reinforcement learning and deep learning enabling the artificial agents in learning the knowledge and experience from actual data directly. With this motivation, this paper presents an Intelligent Video Anomaly Detection and Classification using Faster RCNN with Deep Reinforcement Learning Model, called IVADC-FDRL model. The presented IVADC-FDRL model operates on two major stages namely anomaly detection and classification. Firstly, Faster RCNN model is applied as an object detector with Residual Network as a baseline model, which detects the anomalies as objects. Besides, deep Q-learning (DQL) based DRL model is employed for the classification of detected anomalies. In order to validate the effective anomaly detection and classification performance of the IVADC-FDRL model, an extensive set of experimentations were carried out on the benchmark UCSD anomaly dataset. The experimental results showcased the better performance of the IVADC-FDRL model over the other compared methods with the maximum accuracy of 98.50% and 94.80% on the applied Test004 and Test007 dataset respectively.
Automatic recognition of an acoustic signature in underwater environments is an important and act... more Automatic recognition of an acoustic signature in underwater environments is an important and active field with multiple applications, one of which is vessel recognition. When a vessel moves through the sea, its engine and the cavitation generated by its propellers produce an acoustic wave of unique characteristics that allow for its individual identification. The problem of identification involves several variables, such as ambient noise, biological noise, and even noise produced by its own machinery, which means that the signal produced, is complex to treat. This paper presents a method based on Fourier transform and digital signal processing to extract a set of features allowing for automatic ship classification (by type). Computational intelligence techniques such as Artificial Neural Networks (ANNs) and Support Vector Machines (SVMs) are used for the classification stage. Results showed that the vessel recognition system has accuracy close to 92%.
Este trabajo esta relacionado con las anomalias representadas por discontinuidades o huecos prese... more Este trabajo esta relacionado con las anomalias representadas por discontinuidades o huecos presentes en nubes de puntos tridimensionales. Las auto-oclusiones y las propiedades opticas del material del objeto sensado constituyen las fuentes principales de anomalias. El presente estudio fue realizado con el proposito de analizar similitudes o disimilitudes estadisticas de las propiedades geometricas de los contornos generados por estas dos fuentes. Las propiedades analizadas fueron la estimacion de la curvatura y la torsion, para esto se clasificaron los diferentes contornos en dos grupos: los generados por propiedades opticas del material del objeto y los generados por oclusion. Los resultados sugieren que estas fuentes generan contornos de discontinuidades estadisticamente similares, dificultando una discriminacion basada en estas propiedades.
Rev. Avances en Sistemas Informática, 2010
Traffic surveillance systems promise to improve traffic safety and alleviate traffic jams, among ... more Traffic surveillance systems promise to improve traffic safety and alleviate traffic jams, among other things. The continuous increase in processing speed of computers in the last decade has led to an increasing emphasis on research and development of automatic traffic surveillance systems in different countries. However, traffic surveillance systems are still an open research area because they are not in a state of maturity that allows them to operate with complete autonomy in the task of regulating traffic. This article presents a review of automatic traffic surveillance systems based on vehicular image segmentation techniques. We present a classification of these systems according to their purpose and nature, and a classification of the most commonly used techniques in such systems.
Sensors (Basel, Switzerland), 2021
High-resolution 3D scanning devices produce high-density point clouds, which require a large capa... more 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 s...
Rev. Avances en Sistemas Informática, 2009
Este articulo presenta un nuevo metodo de simplificacion de nubes de puntos. El metodo propuesto,... more Este articulo presenta un nuevo metodo de simplificacion de nubes de puntos. El metodo propuesto, a diferencia de otros, no requiere la construccion previa de mallas poligonales y es robusto al ruido y a valores atipicos presentes en los datos. El metodo propuesto se compone principalmente de tres etapas. En la primera etapa, se segmenta la nube de puntos en regiones homogeneas, usando el algoritmo kmeans. En la segunda etapa , se ajusta un plano de regresion de componentes principales robusto al ruido en cada cluster para determinar la tendencia local de los puntos. Finalmente, en la tercera etapa, usando un algoritmo genetico se seleccionan los puntos de cada cluster cuyo plano de regresion de analisis de componentes principales minimice el angulo con el plano de regresion del cluster. Resultados exper imentales muestran que la distribucion local y global de la nube de puntos original se mantiene.
2019 14th Iberian Conference on Information Systems and Technologies (CISTI), 2019
The human brain and the human vision system, in front of a given scene, focuses on regions with m... more The human brain and the human vision system, in front of a given scene, focuses on regions with more information. Many pieces of research in the field of psychology, neuropsychology and cognitive neurosciences, have detailed the way in which the extraction of such information is carried out, even proposing models of the functioning of visual attention. From these models, computer scientists have proposed computational variants, which imitate human visual attention. This field of research is known as saliency detection on images. In this paper, a new and simple approach for detecting visual saliencies based on Dictionary Learning and Sparse Coding is presented. Our method first subdivides the image into full overlapped patches and runs a dictionary learning over them for obtaining its sparse representation. Then, by analyzing the sparse coding matrix, we compute how many image patches a dictionary atom affects in order to classify them as frequent or rare. Then, we calculate the sali...
Different factors affect the objects that are considered cultural heritage of the regions, among ... more Different factors affect the objects that are considered cultural heritage of the regions, among the most common causes we can find the environmental conditions, long exposure to light, fungi presence, among others. The cultural heritage objects conservation is desirable from the regional cultural importance point of view, therefore, is interesting to explore new technological-based techniques addressed to the conservation. This paper presents an application case of archeological pieces three-dimensional reconstruction using scanner based on structuredlight. It describes the whole procedure including the stages of data acquisition, preprocessing, registration, anomalies correction and visualization. These stages were making using open software, thus providing a precise and cost effective solution. The described setup allows to reproduce digitally an accurate three-dimensional model without need specialized software and hardware.
Automatic recognition of an acoustic signature in underwater environments is an important and act... more Automatic recognition of an acoustic signature in underwater environments is an important and active field with multiple applications, one of which is vessel recognition. When a vessel moves through the sea, its engine and the cavitation generated by its propellers produce an acoustic wave of unique characteristics that allow for its individual identification. The problem of identification involves several variables, such as ambient noise, biological noise, and even noise produced by its own machinery, which means that the signal produced, is complex to treat. This paper presents a method based on Fourier transform and digital signal processing to extract a set of features allowing for automatic ship classification (by type). Computational intelligence techniques such as Artificial Neural Networks (ANNs) and Support Vector Machines (SVMs) are used for the classification stage. Results showed that the vessel recognition system has accuracy close to 92%.
Rev. Avances en Sistemas Informática, 2009
In this paper is presented a method to surface segmentation from sampled points, without a previo... more In this paper is presented a method to surface segmentation from sampled points, without a previous surface approximation, like triangular mesh or local polynomial regression. The segmentation process proceed from the information give by the data points, by this reason are necessary use algorithms, that estimates geometric properties like normals, curvatures and detection of surface discontinuities without its approximations. This peper presents a method to estimate the normals to each point in the surface, taking in account the local
Recently, intelligent video surveillance applications have become essential in public security by... more Recently, intelligent video surveillance applications have become essential in public security by the use of computer vision technologies to investigate and understand long video streams. Anomaly detection and classification are considered a major element of intelligent video surveillance. The aim of anomaly detection is to automatically determine the existence of abnormalities in a short time period. Deep reinforcement learning (DRL) techniques can be employed for anomaly detection, which integrates the concepts of reinforcement learning and deep learning enabling the artificial agents in learning the knowledge and experience from actual data directly. With this motivation, this paper presents an Intelligent Video Anomaly Detection and Classification using Faster RCNN with Deep Reinforcement Learning Model, called IVADC-FDRL model. The presented IVADC-FDRL model operates on two major stages namely anomaly detection and classification. Firstly, Faster RCNN model is applied as an obj...
Advances in data acquisition technology have made it possible to obtain high-density samples from... more 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.
Sensors
Magnetic Resonance (MR) Imaging is a diagnostic technique that produces noisy images, which must ... more Magnetic Resonance (MR) Imaging is a diagnostic technique that produces noisy images, which must be filtered before processing to prevent diagnostic errors. However, filtering the noise while keeping fine details is a difficult task. This paper presents a method, based on sparse representations and singular value decomposition (SVD), for non-locally denoising MR images. The proposed method prevents blurring, artifacts, and residual noise. Our method is composed of three stages. The first stage divides the image into sub-volumes, to obtain its sparse representation, by using the KSVD algorithm. Then, the global influence of the dictionary atoms is computed to upgrade the dictionary and obtain a better reconstruction of the sub-volumes. In the second stage, based on the sparse representation, the noise-free sub-volume is estimated using a non-local approach and SVD. The noise-free voxel is reconstructed by aggregating the overlapped voxels according to the rarity of the sub-volumes it...
Ingenieria y Universidad
Introduction: The use of mobile applications has increased in the last years. Most of them requir... more Introduction: The use of mobile applications has increased in the last years. Most of them require the knowledge of the user location, either for their core service or for marketing purposes. Location-based services (LBS) offer context-based assistance to users based on their location. Although these applications ask the user for permission to use their location and even explain in detail how this information will be used in its terms and conditions, most users are not aware or even interested in the fact that their location information is stored in databases and monetized by selling it to third-party companies. Regarding this situation, we developed a study with the aim to assess perception, concerns and awareness from users about their location information. Methods: This work is based on an exploratory survey applied to the university community, mainly from the North Coast of Colombia, to measure the perception of location privacy of users with mobile devices. The questionnaire wa...