Informática

URI Permanente desta comunidade

http://repositorio.ufes.br/handle/10/17625

Programa de Pós-Graduação em Informática

Centro: CT
Telefone: (27) 4009 2324 R*5126
URL do programa: http://www.informatica.ufes.br/pt-br/pos-graduacao/PPGI

Navegar

Agora exibindo 1 - 11 de 11

Algoritmo A-Estrela de estado híbrido aplicado à navegação autônoma de veículos
(Universidade Federal do Espírito Santo, 2013-08-28) Gonçalves, Michael André; Santos, Thiago Oliveira dos; Souza, Alberto Ferreira de; Gonçalves, Claudine Santos Badue; Aguiar, Edilson de; Chaimowicz, Luiz
In this work, we investigated the use of A-star algorithms (A*) with hybrid state in au-tonomous navigation of vehicles in a three-dimensional space. We have modeled the vehicle position (origin), the goal and other points of interest in the world (states) as nodes of a graph. The cost of navigating between these nodes were modeled as edges of the graph, and a variant of the A* algorithm was used to choose the best path be-tween origin and goal. In order to be able to avoid obstacles and achieve fast algorithm, we used a combination of two heuristics to estimate the cost of the current node to the goal node: one considering only the obstacles and without the limitation of rotation of the vehicle, and its dual disregarding the obstacles and with limited cinematic R³. We implemented the proposed navigation solution and incorporated it to the framework of robotics CARMEN as a navigation module for autonomous vehicles. Our module interacts with other existing modules (interface modules with sensors, mapping, local-ization, etc.) by means of message exchanging. It enables practical use of the algo-rithm. Results of experiments performed on IARA (Intelligent Robotic Autonomous Automobile - autonomous drive car developed in UFES) showed the viability of using the algorithm in simple and structured environments, such as roads, as well as in un-structured and complex environments, such as parking lots and unpaved areas,
Appearance-based global localization with a hybrid weightless-weighted neural network approach
(Universidade Federal do Espírito Santo, 2018-02-02) Silva, Avelino Forechi; Santos, Thiago Oliveira dos; Souza, Alberto Ferreira de; Oliveira, Elias Silva de; Gonçalves, Claudine Santos Badue; Aguiar, Edilson de; Ciarelli, Patrick Marques
Currently, self-driving cars rely greatly on the Global Positioning System (GPS) infrastructure, albeit there is an increasing demand for global localization alternative methods in GPS-denied environments. One of them is known as appearance-based global localization, which associates images of places with their corresponding position. This is very appealing regarding the great number of geotagged photos publicly available and the ubiquitous devices fitted with ultra-high-resolution cameras, motion sensors and multicore processors nowadays. The appearance-based global localization can be devised in topological or metric solution regarding whether it is modelled as a classification or regression problem, respectively. The topological common approaches to solve the global localization problem often involve solutions in the spatial dimension and less frequent in the temporal dimension, but not both simultaneously. It was proposed an integrated spatio-temporal solution based on an ensemble of kNN classifiers, where each classifier uses the Dynamic Time Warping (DTW) and the Hamming distance to compare binary features extracted from sequences of images. Each base learner is fed with its own binary set of features extracted from images. The solution was designed to solve the global localization problem in two phases: mapping and localization. During mapping, it is trained with a sequence of images and associated locations that represents episodes experienced by a robot. During localization, it receives subsequences of images of the same environment and compares them to its previous experienced episodes, trying to recollect the most similar “experience” in time and space at once. Then, the system outputs the positions where it “believes” these images were captured. Although the method is fast to train, it scales linearly with the number of training samples in order to compute the Hamming distance and compare it against the test samples. Often, while building a map, one collects high correlated and redundant data around the environment of interest. Some reasons are due to the use of high frequency sensors or to the case of repeating trajectories. This extra data would carry an undesired burden on memory and runtime performance during test if not treated appropriately during the mapping phase. To tackle this problem, it is employed a clustering algorithm to compress the network’s memory after mapping. For large scale environments, it is combined the clustering algorithms with a multi hashing data structure seeking the best compromise between classification accuracy, runtime performance and memory usage. So far, this encompasses solely the topological solution part for the global localization problem, which is not precise enough for autonomous cars operation. Instead of just recognizing places and outputting an associated pose, it is desired that a global localization system regresses a pose given a current image of a place. But, inferring poses for city-scale scenes is unfeasible at least for decimetric precision. The proposed approach to tackle this problem is as follows: first take a live image from the camera and use the localization system aforementioned to return the image-pose pair most similar to a topological database built as before in the mapping phase. And then, given the live and mapped images, a visual localization system outputs the relative pose between those images. To solve the relative camera pose estimation problem, it is trained a Convolutional Neural Network (CNN) to take as input two separated images in time and space in order to output a 6 Degree of Freedom (DoF) pose vector, representing the relative position and orientation between the input images. In conjunction, both systems solve the global localization problem using topological and metric information to approximate the actual robot pose. The proposed hybrid weightless-weighted neural network approach is naturally combined in a way that the output of one system is the input to the other producing competitive results for the Global Localization task. The full approach is compared against a Real Time Kinematic GPS system and a Visual Simultaneous Localization and Mapping (SLAM) system. Experimental results show that the proposed combined approach is able to correctly global localize an autonomous vehicle 90% of the time with a mean error of 1.20m compared to 1.12m of the Visual SLAM system and 0.37m of the GPS, 89% of the time.
Facial expression recognition using deep learning - convolutional neural network
(Universidade Federal do Espírito Santo, 2016-03-03) Lopes, André Teixeira; Aguiar, Edilson de; Santos, Thiago Oliveira dos; Goldenstein, Siome Klein; Souza, Alberto Ferreira de
Facial expression recognition has been an active research area in the past ten years, with growing application areas such avatar animation, neuromarketing and sociable robots. The recognition of facial expressions is not an easy problem for machine learning methods, since people can vary signi cantly in the way that they show their expressions. Even images of the same person in one expression can vary in brightness, background and position. Hence, facial expression recognition is still a challenging problem. To address these problems, in this work we propose a facial expression recognition system that uses Convolutional Neural Networks. Data augmentation and di erent preprocessing steps were studied together with various Convolutional Neural Networks architectures. The data augmentation and pre-processing steps were used to help the network on the feature selection. Experiments were carried out with three largely used databases (Cohn-Kanade, JAFFE, and BU3DFE) and cross-database validations (i.e. training in one database and test in another) were also performed. The proposed approach has shown to be very e ective, improving the state-of-the-art results in the literature and allowing real time facial expression recognition with standard PC computers.
Identificação e localização dos limites da região trafegável para a navegação de um veículo autônomo
(Universidade Federal do Espírito Santo, 2012-12-10) Azevedo, Vitor Barbirato; Gonçalves, Claudine Santos Badue; Souza, Alberto Ferreira de; Aguiar, Edilson de; Chaimowicz, Luiz
This work presents a computer vision system for real-time mapping of the traversable region ahead of an autonomous vehicle that relies mostly on a stereo camera. For each new image pair captured by the stereo camera, the system first determines the camera position with re-spect to the ground plane using stereo vision algorithms and probabilistic methods, and then projects the camera raw image to the world coordinate system using inverse perspective map-ping. After that, the system classifies the pixels of the inverse perspective map as traversable or not traversable, generating an instantaneous occupancy map. Next, the instantaneous occu-pancy map is integrated to a probabilistic occupancy map, using the Bayes filter to estimate the occupancy probability of each map location. Finally, the probabilistic occupancy map is used to localize the lateral limits of the traversable region. The performance of the system for mapping the traversable region was evaluated in compari-son to manually classified images. The experimental results show that the system is able to correctly map up to 92.22 % of the traversable locations between 20 and 35 m ahead of an autonomous vehicle with a False Acceptance Rate not superior to 3.57 % when mapping par-allelepiped pavements.
Movimentos sacádicos virtuais baseados em VG-RAM na detecção automática de placas de trânsito
(Universidade Federal do Espírito Santo, 2013-08-29) Fontana, Cayo Magno da Cruz; Aguiar, Edilson de; Souza, Alberto Ferreira de; França, Felipe Maia Galvão; Santos, Thiago Oliveira dos
The task of detecting and recognizing road signs in real environments have been widely researched in recent years. Recently, the number of vehicles on urban roads has grown exponentially. Big problems in these pathways have emerged a result of this growth. Statistics of the United Nations (UN), points traffic accidents as a leading cause of death in the world. With the aim of assisting drivers in the task of detecting and recognizing road signs to alert them about possible changes in the way, or even act to control the car, we present in this dissertation a biologically inspired approach to detect traffic signs based on a Virtual Generalizing Random Access Memory Weightless Neural Networks - VG-RAM WNN. VG-RAM WNN are effective machine learning tools that offer simple implementation and fast training and test. Our VG-RAM WNN architecture models the saccadic eye movement system and the transformations suffered by the images captured by the eyes from the retina to the superior colliculus in the mammalian brain. We evaluated the performance of our VG-RAM WNN system on traffic sign detection using the German Traffic Sign Detection Benchmark (GTSDB). Using only 12 traffic sign images for training, our system was ranked in the 16th position, in the total 53 methods submitted among 18 teams, for the prohibitory category in the German Traffic Sign Detection Competition, part of the IJCNN 2013. Our experimental results showed that our approach is capable of reliably and efficiently detect a large variety of traffic sign categories using a few training samples.
Planejamento de movimento para veículos convencionais usando Rapidly-Exploring Random Tree
(Universidade Federal do Espírito Santo, 2013-08-26) Radaelli, Rômulo Ramos; Souza, Alberto Ferreira de; Gonçalves, Claudine Santos Badue; Wolf, Denis Fernando; Aguiar, Edilson de
This work presents a motion planning for car-like robots based on the Rapidly-Exploring Random Tree (RRT) algorithm. RRT was chosen because it is capable of handling with the motion model of a car-like vehicle that is subject to kinematic and dynamic constraints, which make the motion planning problem harder. Kinematic constraints of a car-like robot prevent it from turning around its own axis or moving laterally, and dynamic constraints limit its speed and acceleration. The motion planner developed in this work is able to plan on the road and in open areas. When planning on the road, the goal is to keep the car in its lane, without invading the counter hand. When planning in open areas, the goal is to generate trajectories able to park the car. The performance of the motion planner was evaluated in a simulation framework for autonomous vehicles and in a robotic platform called IARA. IARA is built upon the Ford Escape Hybrid equipped with sensors and modified with mechanisms for controlling the throttle, brake, steering wheel position, etc., through computers installed in the car. Experimental results demonstrate that the motion planner is capable of planning collision-free trajectories in real time. Moreover, the trajectories generated by the motion planner can spend an amount of energy equivalent to those generated by a driver.
Sistema de rastreamento visual de objetos baseado em movimentos oculares sacádicos
(Universidade Federal do Espírito Santo, 2015-04-09) Andrade, Mariella Berger; Santos, Thiago Oliveira dos; Souza, Alberto Ferreira de; Gonçalves, Claudine Santos Badue; Aguiar, Edilson de; Salles, Evandro; França, Felipe Maia Galvão
Visual search is the mechanism that involves a scan of the visual field in order to find an object of interest. The brain region responsible for performing the visual search, performed by saccadic eye movements, is the Superior Colliculus. A computer system for visual search biologically inspired needs to modell the saccadic eye movement, the transformation suffered by the images captured by the eyes in the way from the retina to the Superior Colliculus, and the response of the neurons of the Superior Colliculus to patterns of interest in the visual scene. In this work, we present a biologically inspired long-term object tracking system based on Virtual Generalizing Random Access Memory (VG-RAM) Weightless Neural Networks (WNN). VG-RAM WNN is an effective machine learning technique that offers simple implementation and fast training. Our system models the biological saccadic eye movement, the transformation suffered by the images captured by the eyes from the retina to the Superior Colliculus (SC), and the response of SC neurons to previously seen patterns. We evaluated the performance of our system using a well-known visual tracking database. Our experimental results show that our approach is capable of reliably and efficiently track an object of interest in a video with accuracy equivalent or superior to related work.
Sistema GoPro estéreo submarino de reconstrução 3D
(Universidade Federal do Espírito Santo, 2016-09-30) Guimarães, Juliana Amorim; Santos, Thiago Oliveira dos; Aguiar, Edilson de; Souza, Alberto Ferreira de; Komati, Karin Satie; Andrade, Mariella Berger
abstract
Um sistema de realidade aumentada para a visualização de objetos virtuais e faciais utilizando câmeras RGB-D
(Universidade Federal do Espírito Santo, 2016-09-23) Azevedo, Pedro Henrique Vieira de Oliveira; Santos, Thiago Oliveira dos; Aguiar, Edilson de; Souza, Alberto Ferreira de; Giraldi, Gilson Antônio
This work presents a virtual try-on system to correctly visualize 3D objects (e.g., glasses) in the face of a given user. By capturing the image and depth information of a user through a low-cost RGB-D camera, we apply a face tracking technique to detect specific landmarks in the facial image. These landmarks and the point cloud reconstructed from the depth information are combined to optimize a 3D facial morphable model that fits as good as possible to the user’s head and face. At the end, we deform the chosen 3D objects from its rest shape to a deformed shape matching the specific facial shape of the user. The last step projects and renders the 3D object into the original image, with enhanced precision and in proper scale, showing the selected object in the user’s face. We validate the performance of our system numerically and visually based on a facial database composed by images and point clouds of male and female subjects with different ages and face shapes.
Uma proposta de correção semiautomática de questões discursivas e de visualização de atividades para apoio à atuação do docente
(Universidade Federal do Espírito Santo, 2014-09-25) Pissinati, Eduardo Marim; Oliveira, Márcia Gonçalves de; Oliveira, Elias Silva de; Nobre, Isaura Alcina Martins; Aguiar, Edilson de
The task of correcting discursive texts requires much effort and dedication from teachers. The larger the number of students, the bigger the work is, until it reaches a point where the teacher fails to apply these activities since he cannot evaluate them, subjecting his students only to objective questions. Furthermore, the identification of students who need more attention is impaired in very large groups. Increasing the number of teachers per class could solve these problems, but more teachers represent more costs to the institution. In this context, the use of the computer can assist in these activities without including new teachers. However this strategy must be well thought not to overload the teacher with inadequate tools. This paper presents a tool for semi-automatic correction of discursive texts and a visualization tool that will allow the teacher to monitor individual performances of their students even in large classes. The semi-automatic correction tool uses intelligent recovering techniques of the information as clustering to correct discursive answers. Besides that, the visualization tool uses information visualization techniques to present the results to the teacher. Both were integrated with Moodle Virtual Learning Environment in order not to change the normal routine of work of those who use it.
Vision-based ego-lane analysis system: dataset and algorithms
(Universidade Federal do Espírito Santo, 2016-08-03) Berriel, Rodrigo Ferreira; Aguiar, Edilson de; Santos, Thiago Oliveira dos; Souza, Alberto Ferreira de; Schwartz, William Robson
Lane detection and analysis are important and challenging tasks in advanced driver assistance systems and autonomous driving. These tasks are required in order to help autonomous and semi-autonomous vehicles to operate safely. Decreasing costs of vision sensors and advances in embedded hardware boosted lane related research – detection, estimation, tracking, etc. – in the past two decades. The interest in this topic has increased even more with the demand for advanced driver assistance systems (ADAS) and self-driving cars. Although extensively studied independently, there is still need for studies that propose a combined solution for the multiple problems related to the ego-lane, such as lane departure warning (LDW), lane change detection, lane marking type (LMT) classification, road markings detection and classification, and detection of adjacent lanes presence. This work proposes a real-time Ego-Lane Analysis System (ELAS) capable of estimating ego-lane position, classifying LMTs and road markings, performing LDW and detecting lane change events. The proposed vision-based system works on a temporal sequence of images. Lane marking features are extracted in perspective and Inverse Perspective Mapping (IPM) images that are combined to increase robustness. The final estimated lane is modeled as a spline using a combination of methods (Hough lines, Kalman filter and Particle filter). Based on the estimated lane, all other events are detected. Moreover, the proposed system was integrated for experimentation into an autonomous car that is being developed by the High Performance Computing Laboratory of the Universidade Federal do Espírito Santo. To validate the proposed algorithms and cover the lack of lane datasets in the literature, a new dataset with more than 20 different scenes (in more than 15,000 frames) and considering a variety of scenarios (urban road, highways, traffic, shadows, etc.) was created. The dataset was manually annotated and made publicly available to enable evaluation of several events that are of interest for the research community (i.e. lane estimation, change, and centering; road markings; intersections; LMTs; crosswalks and adjacent lanes). Furthermore, the system was also validated qualitatively based on the integration with the autonomous vehicle. ELAS achieved high detection rates in all real-world events and proved to be ready for real-time applications.

Navegar

Navegando Informática por Autor "Aguiar, Edilson de"

Resultados por página

Opções de Ordenação