Accepted Works

Uma câmera inteligente para reconhecimento automático de frame markers

Projeto QualisFruit

The agricultural hub of Petrolina and Juazeiro is considered the most dynamic pole of the backlands of Pernambuco and Bahia, and are distinguished by high export of mango, especially the kind of "Tommy Atkins".

However, one of the great difficulties of producers is to determine the point of harvest, since fruits do not ripen prematurely harvested or are irregularly shaped and this is reflected in the amount of sugars and acidity of mango. On the other hand, a late harvest reduces the lifetime and makes the fruit more sensitive to mechanical damage and attack by micro- organisms.

Thus, to meet the demand for export of fruits, one should ensure the quality of the fruits that are harvested in this region. One tool that can assist in this task is the digital image processing, where the fruits may have set the point of harvest, based on analysis of their appearance to meet the market it is intended for.

In this work a system for detecting the point of harvest mango "Tommy Atkins" through methods of converting the RGB (Red, Green, Blue) and HSV (hue, saturation, value) is presented, using only integers and being able to detect the stage of ripeness second criterion developed by Embrapa.
The current project will eventually operate in remote monitoring of orchards, with the aid of Unmanned Aerial Vehicles (UAV), called drones, and, given his analysis performed in hardware, and software featuring more processing capacity to integrate new features the system. 

Mapeamento 3D usando LiDAR

Cancelamento Ativo de Ruídos em Ambientes

A equipe pretende desenvolver um sistema de escaneamento 3D baseado em tecnologia LiDAR (Light Detection and Ranging).

Uma abordagem hardware/software para reconhecimento do alfabeto em libras através das mãos.

About 170 thousand people in Brazil have self declared “deaf”. Because they are not able to communicate so clearly as those who does not have the disability, this portion of the population needs a written or visual mean. To cope with this need in Brazil, an official sign language pattern, called LIBRAS, that allows the communication among deaf, was developed. However, still there is a communication gap using this sign language, once only a small part of the population knows LIBRAS. Trying to reduce this gap, this approach proposes a computational solution to translate LIBRAS symbols to text symbols, through computer vision. To guarantee the real-time translation, some hardware components were built to fulfill the most expensive image processing tasks using in this project: RGB color space to YCrCb color space conversion and pixel thresholding. Once real-time videos capture produces a high number of data and the target platform have some limitation in the communication throughput, some adaptation, both in hardware and software, were required. These adaptations involve the compression of the thresholded data, a memory buffer manager and a frame collector with a data decompression software component. A series of system architectures coping with this adaptations were designed and implemented to reach desired throughput. This approach represents a initial stage of a bigger project. Although experimental results has shown that the system recognize real-time 12 LIBRAS symbols with the accuracy of 92.5%, some isolated symbol, which are represented through a sequential movement are not actually able to be detect.

Desenvolvimento de Dispositivo Eletrônico Móvel para Auxiliar o Diagnóstico de Glaucoma de Ângulo Aberto

Glaucoma is the second leading cause of blindness in the world. In 2010, the number of people which had the disease was around 60 million, and by 2020 this number is estimated to increase to 80 million. Besides, if not diagnosed in an early stage, glaucoma can lead to irreversible blindness. About 8.4 million people in the world is blind due to glaucoma and by 2020 this number is estimated to reach 11.2 million. In retinal images there are two regions of interest for the diagnosis of Glaucoma, they are the Optic Disc and the Optic Cup. A common way to detect glaucoma consists of analyzing the proportion between the diameters of the Disc and the Cup (Cup to Disc ratio). A ratio greater than 0.5 is a good indicative of glaucoma. In this project, a prototype of an embedded device that automatically calculates the Cup to Disc ratio from retinal images was developed. A combination of FPGA and the Atom processor was used to reduce the computational costs related to the image processing, and to create an interface and a database where is possible to keep track of the progression of the disease through the years. This portable device, which can be easily used by anyone, helps providing a better medical care in the neediest areas, such as quilombolas and indian communities or regions with poor medical infrastructure, enabling a prior identification of Glaucoma, thus favoring a better life quality for all population.

Automatic Interpreter of the Brazilian Sign Language

Guia Inteligente móvel para deficientes visuais

The everyday task of navigating one’s surrounding can amount to a true challenge to the visually impaired. Buildings are usually not designed with the challenges of this particular group of people in mind, which further aggravates the issue. Guide dogs have long been the most common attempt at solving this problem. However, guide dogs are quite limited, for they are incapable of reading signs and other visual clues which humans use naturally, and they can’t perform navigation, having to rely on their owners to also be a guide to them. Furthermore, the cost and effort involved in having a guide is often an issue, and the cost and time involved in training a guide dog means that one is not readily obtained. We aim to produce a system that is both lower in cost and richer in capabilities than the traditional solution to the problem, namely, guide dogs, while also being easier to obtain, and intend to achieve this via the recent advances in the processing power of embedded systems, the ease of access and use of a range of sensors, and the ubiquity of network connectivity. We propose our solution in the form a robotic car capable of guiding its user around buildings. The car is equipped with sensors that allow it to follow a path, avoid obstacles and determine when it has reached a destination. The system will be also capable of taking in voice commands, which in turn removes another barrier to the visually impaired.

Sistema de detecção de obstáculos via eco localização.

O projeto consiste utilizar um sistema de localização de objetos a curta distân- cia por meio da reflexão de ondas sonoras em superfícies próximas.Dessa forma, imagi- na-se um ente móvel com finalidade de se localizar espacialmente sem a ajuda de câme- ras ou outros periféricos.

Data Logger for the Public Transport (RodoMonitor)

A otimização no transporte público não inclui apenas uma melhor programação dos horários de circulação e das dimensões dos ônibus a serem utilizados em cada linha e período, mas também na elaboração de trajetos mais inteligentes. Atualmente, o Plano de Mobilidade Urbana Sustentável da Grande Florianópolis (PLAMUS) está realizando um estudo minucioso, ainda que manual, para identificar os “gargalos” do transporte público na cidade de Florianópolis1 para então propor soluções no sistema viário da capital. Neste sentido, esta proposta visa desenvolver um sistema embarcado (RodoMonitor) com o objetivo de realizar esta coleta de dados de forma automatizada, inteligente e direcio- nada de dados no transporte público da cidade de Florianópolis em Santa Catarina. 

Monitoring System for the Elderly Fall

O projeto consiste em monitorar queda de idosos em residências e ambientes críticos e propícios a quedas, através de câmeras, processamento e tratamento de imagens. Será identificado situações que indiquem movimentos de quedas e tombos. Após o reconhecimento e a interpretação da queda, serão enviadas notificações para os familiares, assim como para instituições de atendimento de emergência. As notificações serão via celular, e-mail e telefones residenciais e serão enviadas automaticamente e instantaneamente diante do reconhecimento da queda.

Sistema de Identificação Biométrica de Indivíduos Através da Antropometria e do Caminhar Humano

Identifying individuals using biometric data is an important task in surveillance, authentication and even entertainment. This task is even more challenging when required to be performed without physical contact and at a distance. The most recent trend is the use of full-body measurements to perform passive identification. The measurement of body parts (anthropometry) and gait analysis are examples of techniques that leverage such approach. In this work, the Microsoft Kinect and its NUI Skeleton API are used to extract automatically a skeleton from the user or users standing in front of the sensor. Such automatic extraction has the potential to simplify the process of extracting useful anthropometric and gait features for biometric identification. Thus, as the DE2i-150 Development Kit contains a processor that operates at a frequency below the required one by the Kinect sensor, the gait data cannot be captured, just anthropometric attributes. However, anthropometric attributes are sufficient to gender identification, as we do in this work. The process work as follows: the Microsoft Kinect sensor is connected to the DE2i-150 Development Board and captures video footages from a person. This person has his/her anthropometric attributes captured and the Euclidian distance is calculated in order to take the measures for all main parts of the body. This information is processed by a neural network, which output indicates if it is a male or female person. A storefront is also part of the application and shows female or male products, depending on the gender identification system’s output.

Li-Fi connected Media Center

A placa fornecida pela Intel será utilizada como servidor multimídia para ambientes fechados, como residências. Este servidor guardará todos os dados de áudio, vídeo e afins e estará conectado às lâmpadas do ambiente, de forma a prover conteúdo aos clientes através de Li-Fi. Os clientes serão placas (como a Raspberry Pi) conectadas a dispositivos de reprodução (e.g. televisão, som) e a receptores Li-Fi, permitindo que um arquivo transmitido pelo servidor seja captável e reprodutível simultaneamente por todos os dispositivos, resultando em um baixo consumo energético total.

Project Annuit Walk - PAW

O Projeto PAW trata-se de um óculos inteligente para deficientes visuais com foco na detecção de obstáculos. A ideia principal é promover a inclusão social dessas pessoas, dando mais conforto e segurança na sua mobilidade. Nossa pretensão é trazer uma experiência de cidade mais acessível para uma melhora em sua qualidade de vida. Além disso, para usuários portadores de smartphones, pretendemos também coletar dados referentes a qualidade das ruas por onde passam, calcular o nível de acessibilidade dessas áreas e notificar os órgãos responsáveis com o objetivo de melhorar a acessibilidade da sua região.

Equipamento para medição e caracterização de canais e ruídos PLC

Nowadays, further than the common usage of the PL channel (Power line channel) on electric power transmission, it can also be used to transmit data over the same structure. That data transmission is called PLC (Power line Communication). The PLC kind of communication inherit several issues due to the instability of the channel. The noise and the channel behavior, without treatment, disrupts the correct data transmission. Sometimes, the channel fading and the noise are so strong that the proper communication becomes infeasible. The SNR (Signal to Noise Ratio) is a parameter that can be used to analyze the channel behavior and capacity. The first version of i9Analytics is a set of software and hardware that is able to make the analysis of a PLC channel between 1.7 and 50 MHz considering the values of SNR. The project consists of two transceivers; the first one is an OFDM (Orthogonal Frequency Division Multiplex) communication system transmitter, implemented on a development kit DE2-115. The second is a receiver, implemented on the development kit Intel DE2-150i, that analyzes the signal fading and extract the channel parameters. The communications occurs in the PLC channel due to the front-end developed by the team. It has analogic/digital and digital/analogic converters, a gain chip to amplify or attenuate the transmitted and received signal and analogic filters to attenuate the power signal.