1 Figure of Merit: Performance Ratio • Suppose we are working on an optimization problem in which each potential solution has a positive cost, and we want to find one of near-optimal cost. When the objective is to minimize the cost, the near-optimal is always going to be more than the optimal. When the objective is to maximize profit, the near-optimal is less that the optimal. • The standard measure (figure of merit) in Theory of Algorithms to measure the quality of approximation is the ratio:
As Wireless Mesh Networks (WMNs) are growing day-by-day, a seamless and secure handoff is gaining significant importance for supporting multi-hop WMNs. On this point, various authentication protocols, such as the ticket-based handoff authentication, for wireless mesh networks have been proposed. Modeling and formal verification of the aforementioned protocol using CPN Tools and ASK-CTL statement are the purposes of this paper. To this aim, the resistance of the protocol against attacks, such as the man in the middle attack is investigated and then, it is concluded that it is secure.
The technique of image recognitions is becoming more and more important to identify objects, places, and people. Currently, several deep learning methods on image recognition have been proposed. To identify multiple targets, the notion of region proposal has proposed which uses multiple resolution methods to improve accuracy, such as R-CNN, Fast R-CNN, Faster R-CNN, Mask R-CNN, SSD, and YOLO. However, these improvements are based on pixel. Still, there are uncertain objects which the human eye observes in the surrounding scene. At this time, we make guesses based on other, more clear objects. In the paper, we propose a method for object recognition using the probability of correlation between the objects. When performing object recognition in an image, we calculate the probability of correlation between the objects to adjust the related parameters and the weight values. Our proposed method improve the overall recognition of objects in the image.
The association of correct object locations to its respective track over time is vital for robust tracking results when employing a detection-based tracking approach. It is important for handling object track identity switches which is one of the limiting issues when designing multiple-object trackers. This paper expresses an object model by its saliency-colour histogram. Dissimilarity between a reference track model and input candidate detection (obtained by an object detector) is quantified by a measure derived from the Bhattacharyya coefficient. The Bhattacharyya distance measure is used for validating the confidence in tracks based on an adaptively obtained threshold value. This approach results in improved tracker results under a track-identity loss scenario, especially under short-term partial or full occlusions, clutter or scale variations. The suggested method for track assignment, combined with a state-of-art real-time object detector YOLO (i.e. “You only look once”) led to improved online results by reducing the number of miss-assignments within tracks.
In portfolio management, stock selection and evaluation can be based on a variety of financial attributes over a period of time. It has been shown recently by Irukulapati et al. that long term portfolio management strategy using attribute selection and combinatorial fusion can not only achieve better results than individual attributes but also exceed the performance of the Russell 2000 index. In this paper, we propose a method to compute the attribute scoring system using weighted average by recency (AR) giving more weight to scores at the time closer to the present. We then show, by market testing, that our results perform better than that of Irukulapati et al. in a majority of cases as well as the Russell 2000.
Massive Machine-Type-Communication (mMTC) is expected to play a crucial role in 5G networks to enable Internet of Things (IoT). But with deployment of lots of mMTC devices, mobile cellular network will suffer from problems of congestion and large system overhead in both the Radio Access Network (RAN) and Core Network (CN). Currently multiple proposals, such as extended access barring (EAB) and access class barring (ACB), have been broadly discussed in Third Generation Partnership Project (3GPP) to combat the problem of Random Access Channel (RACH) congestion. However, less effort has been put on the efficiency issue of uplink transmission for mMTC traffics featured with small data packets and infrequent transmissions. To address this problem, we present an enahnced grant-free access scheme for mMTC uplink transmissions based on the probability concept, where a type of specific resource called Probability-Based Access (PBA) channel is allocated with congestion probability indicated. Thus the mMTC device can initiate the uplink transmission based on the probability, i.e. data can be transferred directly on the PBA channel, or fall back to the legacy procedure by using contention-based random access scheme. The performance of the proposed scheme is evaluated by numerical simulations and its effectiveness and advantages are validated.
Utilizing emerging information technology in agriculture automation is arisen for reducing human errors and increasing the productivity and quality. This paper proposes a simple algorithm OCG with deep learning network to determine the grading levels of Oncidium orchid cut flowers which are related to the sale prices. The algorithm consists of two phases. The grading criteria about lengths are estimated by image analysis in the first phase, while the grading criteria about counting branches are predicted using deep learning in the second phase. The experimental results show that our algorithm can achieve accuracy of 0.8 and the algorithm is practical.
With the development of new medical auxiliaries such as virtual reality and surgery robotics, recommender systems are emerged to interact with the medical auxiliaries and support doctor’s decisions and operations, especially in collaborative healthcare, recommender systems can interactively take into account the preferences and concerns from both patients and doctors. However, how to apply and integrate recommender systems is still not clear in collaborative healthcare. Therefore, from practical perspective this paper investigates the application of recommender systems in three typical collaborative healthcare domains, which are augmented/virtual reality, medicine and surgery robotics. The results not only provide the insights of how to integrate recommender systems with healthcare auxiliaries but also discuss the practical guidance of how to design recommender systems in collaborative healthcare.
Taking full account of the advantages and disadvantages of monocular vision and binocular vision in environmental perception, this paper proposes a single-binocular vision conversion strategy based on monocular camera for obstacle detection at non-signalized intersections. Using this strategy we can get fast and accurate identification of obstacles on the lateral anterior of the vehicle, and achieve fast and accurate detection for subsequent distances and speeds of obstacles. To ensure driving safety at the intersections on the premise of effectively detecting cross-conflict, first, the monocular camera conversion model, the number of cameras and the installation position are determined. Second, a single-binocular vision conversion strategy is made: the single-binocular visual composition is determined in real time according to the distance parameter and the angle parameter, and the obstacle information on the lateral anterior of the vehicle is acquired. In the end, compared with the complete monocular vision and complete binocular visual environmental perception method, the comparison results prove the rationality of the conversion strategy, and prove that the strategy is efficient and accurate in environmental perception and obstacle recognition, ranging, and speed measurement.
Authentication is one of the most important issues in the information technology field and, primarily, in Cryptography. With the diffusion of devices for the Internet of things, the interest in efficient and with low computational loads authentication protocols, has increased more and more in the last few years. Indeed, traditional protocols, based on symmetric primitives and, in general, on pseudo-random functions, are not suitable for computationally constrained devices. On the other hand, one of the most interesting families of protocols for such devices seems to be the one based on the learning parity with noise problem. In this paper, building on some previous works, we propose a new instance of a lightweight authentication protocol constructed on this problem. We describe some optimizations which could be employed to improve the efficiency of an implementation, and we give also a look at the real world, discussing the applicability of our proposal to several devices available on the market.
Ship movement information is becoming increasingly available, resulting in an overwhelming increase of data transmitted to human operators. Understanding the Maritime traffic patterns is important to Maritime Situational Awareness (MSA) applications in particular, to classify and predict trajectories on sea. Therefore, there is need for automatic processing to synthesize the behavior of interest in a simplified, clear, and effective way without any loss of data originality. In this paper, we propose a method to calculate route prediction from a synthetic route representation data once the picture of the maritime traffic is constructed. The synthetic route knowledge based on Automatic Identification System (AIS) is used to classify and predict future routes along which a vessel is going to move. This is in agreement with the partially observed track and given the vessel static and dynamic information. The prediction results do not only reduce data storage space in the database but can also supply data support for traffic management, accident detection, and avoidance of automatic collision and therefore promote the development of maritime intelligent traffic systems. Finally, the simulation results shows a good tradeoff between the predicted and the actual observed vessel routes.
Aiming at the global path planning of intelligent vehicles, an optimal hybrid path planning algorithm considering the dynamic and static characteristics of intelligent vehicles is proposed. On the grid map with known static information of the environment, the improved A* algorithm is used for global path planning, and the obstacles in the path are expanded according to the static characteristics of the intelligent vehicle itself. Combined with the dynamic characteristics of the intelligent vehicles, dynamic window approach is used to carry out the local obstacle avoidance and path planning of the vehicle according to the unknown and varied environmental information around the vehicle. On this basis, the key turning point in the global path is used as the sub-target point correction of Dynamic Window Approach (DWA). The simulation results show that the proposed method can be used to avoid dynamic and static obstacles by guiding the vehicle to the target ending. Additionally, the dynamic constraints of the vehicle are satisfied during the journey without collision with the road boundary, which ensures the stability and safety of the vehicle.
The integration of semantic web and big data is a key factor in the definition of efficient model to represent knowledge and implement real world applications. In this paper we present a multimedia knowledge base implemented as a semantic multimedia big data storing semantic and linguistic relations between concepts and their multimedia representations. Moreover, we propose a document visualization strategy based on statistical and semantic analysis of textual and visual contents. The proposed approach has been implemented in a tool, called Semantic Tag Cloud, whose task is to show in a concise way the main topic of a document using both textual features and images. We also propose a case study of our approach and an evaluation from a user perception point of view.
Paper addresses security issues with public access WiFi networks, with emphasis on networks deployed in touristic places, because of their popularity. Such networks are often poorly administered and guarded whereas tourist-services they support, are massively used by thousands of occasional users daily. With intention to put emphasis on the security awareness of the users, filed research was conducted to investigate current security preferences of wireless computer networks in tourist destination, in the City of Zadar, Croatia. The research was conducted during preparation and early in the tourist season, spring/summer 2018. Hardware research support include AP beacon used a TL-WN722N card with a data rate of 150 Mbps, a 5 db antenna, a chip Atheros AR9271, all powered by Linux operating. Small suite was a passive scan tool for the beacon area. The data set used include the default AP settings that transmits its current SSID every 100 ms. WLAN card was used in the vehicle that was set up in the monitor mode used to collect all the available beacon frames. In addition to field research, we conduct additional survey with aim to investigate the general habits of users of wireless computer networks, from personal perspective. Overall goal was to put attention on WiFi security awareness and to expose security behaviour at router level.
Traditional cloud computing has the challenge to serve many clients with many services. Spreading the services to across of edge server will reduce the load of traditional cloud computing. Kubernetes is one of the platforms used for cloud management. Kubernetes helps to deploy, and scaling the application. Nowadays, a lot of communities build a lightweight Kubernetes than suitable for edge device such as Raspberry Pi. This paper Investigate the performance of Kubernetes lightweight that installed in the Raspberry Pi.
Recently, Deep Neural Networks have been successfully utilized in many domains; especially in computer vision. Many famous convolutional neural networks, such as VGG, ResNet, Inception, and so forth, are used for image classification, object detection, and so forth. The architecture of these state-of-the-art neural networks has become deeper and complicated than ever. In this paper, we propose a method to solve the problem of large memory requirement in the process of training a model. The experimental result shows that the proposed algorithm is able to reduce the GPU memory significantly.
Recently, containers have been used extensively in the cloud computing field, and several frameworks have been proposed to schedule containers using a scheduling strategy. The main idea of the different scheduling strategies consist to select the most suitable node, from a set of nodes that forms the cloud platform, to execute each new submitted container. The Spread scheduling strategy, used as the default strategy in the Docker Swarmkit container scheduling framework, consists to select, for each new container, the node with the least number of running containers. In this paper, we propose to improve the Spread strategy by presenting a new container scheduling strategy based on the power consumption of heterogeneous cloud nodes. The novelty of our approach consists to make the best compromise that allows to reduce the global power consumption of an heterogeneous cloud infrastructure. The principle of our strategy is based on learning and scheduling steps which are applied each time a new container is submitted by a user. Our proposed strategy is implemented in Go language inside Docker Swarmkit. Experiments demonstrate the potential of our strategy under different scenarios.
Nowadays, plenty of digital services are provided to citizens by means of terminals located in public unguarded places. In order to access the desired service, users, authenticate themselves by providing their credentials through such terminals. This approach opens up to the problem of fraudulent devices that could be installed in place of regular terminals to capture users’ confidential information. Indeed, despite the development of increasingly secure systems aiming at guaranteeing an acceptable security level, users are frequently unable to distinguish between terminals on which security measures are enforced (trusted terminals) and malicious terminals that pretend to be trusted. We deal with this problem by presenting a human-compatible authentication protocol, leveraging Graphical Passwords, helps user to authenticate a terminal before using it. We also present a prototype implementation of this protocol, called TRUST (TRust Unguarded Service Terminals). The usability of our solution has been analyzed by means of a preliminary experimentation.
Wireless sensor networks demands proper means in order to obtain an accurate location of their nodes for a twofold reason: on the one hand, the exchanged data must be spatially meaningful since their content may be unusual if the location of where they have been produced is not associated to them, on the other hand, such networks need efficient routing algorithms where optimal routing decisions must be based on location information. Accuracy is not the only demands for positioning of sensors, but also simplicity and infrastructure independence in order to avoid excessive energy consumption and deployment costs. For these reasons, GPS is not used but the RF technologies are mainly preferred. Based on those technologies, most of the solutions tailored for sensors are designed so as to determine a location based on simple measurements of the signal intensity of the received messages. Despite being able to satisfy the peculiar requirements for localization in sensor networks, those methods have been proved to be particularly inaccurate, due to the unreliability of the adopted measurements upon which location is inferred. This article proposes a novel approach for range-free localization by obtaining intensity measurements at different Power of Transmission levels, using them as inputs for multiple location estimators, and aggregating the outputs of those estimators in order to achieve a more accurate determination of a sensor position. We have implemented our solution on real sensor platforms and performed some experiments in order to show how this simple solution allows halving the localization error and reducing the energy consumption of about 18% with respect to the state-of-the-art algorithms.
