Today's industrial facilities, such as oil refineries, chemical plants, and factories, rely on wired sensor systems to monitor and control the production processes. The deployment and maintenance of such cabled systems is expensive and inflexible. It is, therefore, desirable to replace or augment these systems using wireless technology, which requires us to overcome significant technical challenges. Process automation and control applications are mission-critical and require timely and reliable data delivery, which is difficult to provide in industrial environments with harsh radio environments. In this article, we present the GINSENG system which implements performance control to allow us to use wireless sensor networks for mission-critical applications in industrial environments. GINSENG is a complete system solution that comprises on-node system software, network protocols, and back-end systems with sophisticated data processing capability. GINSENG assumes that a deployment can be carefully planned. A TDMA-based MAC protocol, tailored to the deployment environment, is employed to provide reliable and timely data delivery. Performance debugging components are used to unintrusively monitor the system performance and identify problems as they occur. The article reports on a real-world deployment of GINSENG in an especially challenging environment of an operational oil refinery in Sines, Portugal. We provide experimental results from this deployment and share the experiences gained. These results demonstate the use of GINSENG for sensing and actuation and allow an assessment of its ability to operate within the required performance bounds. We also identify shortcomings that manifested during the evaluation phase, thus giving a useful perspective on the challenges that have to be overcome in these harsh application settings.
We are living in a new communication age, which will radically transform the way we live in our society. A world where anything will be connected to Internet is being created, generating an entirely new dynamic network - The Internet of Things (IoT) - enabling new means of communication between people, things and environment. A suitable architecture for Internet of Things (IoT) demands the implementation of several and distinct technologies that range from computing (e.g. Cloud Computing), communications (e.g. 6LowPAN, 3/4G) to semantic (e.g. data mining). Thus, it is necessary to understand very well all these technologies in order to know which of them is most suitable for a given scenario. Therefore, this paper proposes an IoT architecture for disabled people and intends to identify and describe the most relevant IoT technologies and international standards for the stack of the proposed architecture. In particular, the paper discusses the enabling IoT technologies and its feasibility for people with disabilities. At the end, it presents two use cases that are currently being deployed for this population.
30 June 2014 in International Journal of Database Management Systems
Among the activities affected by visual impairment, navigation plays a fundamental role, since it enables the person to independently move in safety. The heterogeneous environment, easily perceived by visually enabled people, is hardly known by partially sighted people. A challenging task for these people is independent navigation in new spaces/buildings/environments.The environment is usually signaled and labeled with visual marks and signs which are not appropriate for blind persons. With the purpose of balancing the access to services and spaces among all persons, this work proposes an innovative navigation and information system to help the navigation of blind people within new environments (e.g. shopping center, public office building). Based on smartphones and wireless sensors deployed in the environment, we propose an information tracking system for realtime guide blind people (BlindeDroid). It offers guided navigation, answering questions, and providing objective information about places, products and services that are available surrounding the user
Image structures are segmented automatically using deep learning (DL) for analysis and processing. The three most popular base loss functions are cross entropy (crossE), intersect-over-the-union (IoU), and dice. Which should be used, is it useful to consider simple variations, such as modifying formula coefficients? How do characteristics of different image structures influence scores? Taking three different medical image segmentation problems (segmentation of organs in magnetic resonance images (MRI), liver in computer tomography images (CT) and diabetic retinopathy lesions in eye fundus images (EFI)), we quantify loss functions and variations, as well as segmentation scores of different targets. We first describe the limitations of metrics, since loss is a metric, then we describe and test alternatives. Experimentally, we observed that DeeplabV3 outperforms UNet and fully convolutional network (FCN) in all datasets. Dice scored 1 to 6 percentage points (pp) higher than cross entropy over all datasets, IoU improved 0 to 3 pp. Varying formula coefficients improved scores, but the best choices depend on the dataset: compared to crossE, different false positive vs. false negative weights improved MRI by 12 pp, and assigning zero weight to background improved EFI by 6 pp. Multiclass segmentation scored higher than n-uniclass segmentation in MRI by 8 pp. EFI lesions score low compared to more constant structures (e.g., optic disk or even organs), but loss modifications improve those scores significantly 6 to 9 pp. Our conclusions are that dice is best, it is worth assigning 0 weight to class background and to test different weights on false positives and false negatives.