查看更多>>摘要：In delivery of media content involving projections of images on non-planar and deformable surfaces, there is an inevitable problem of geometric distortion which causes visually incorrect images to be produced in the display region. Presented in this paper is a simple and computation efficient method for geometric distortion correction. While the simplicity comes from the use of a single uncalibrated camera to capture the projected image, the computation efficiency comes from the use of a small number of image points along the periphery of the display region to estimate the geometric distortion. In particular, cylindrical surface deformation is used to model the geometric displacements of peripheral image points, and bilinear approximation is used to model the geometric displacements of image points lying between the peripheral image points. Results of tests on static and deformable projection surfaces show this distortion-correction technique improved the normalised geometric similarity measure of distorted displayed images by as much as 31%.
查看更多>>摘要：Mobile application development is one of the most promising domains in the software industry. The rapid growth of hardware and emerging technologies has resulted in a large number of mobile platforms, which constitutes a challenge that developers must face when they build applications for different platforms. The model-based artefacts co-evolution is another challenge when metamodel evolves. This paper introduces a model-driven approach for managing the metamodel evolution in the context of cross-platform mobile application development. Firstly, we propose an MDD approach, based on a generic DSL and a set of code generators to generate platform-specific source code. Secondly, the approach provides a graphical framework: (i) to extend the original metamodel through a set of rules, (ii) to define the mapping between the newly added meta-elements and their corresponding in the target platforms and (iii) to automatically update existing code generators. This paper demonstrates the potentials and limits of our approach through a concrete case study.
查看更多>>摘要：Modulation Transfer Function (MTF) compensation had been widely used in Remote-Sensing (RS) imagery processing for texture feature enhancement and high-frequency details restoration. In this paper, we propose an optical satellite MTF Compensation (MTFC) method for engineering applications and RS data production. The framework mainly includes Point Spread Function (PSF) model estimation and non-blind image deconvolution. In PSF estimation, we used the iterative total variation algorithm and 2D Gaussian model fitting to reduce the dependence on slant-edge quality and improve the efficiency of PSF measurement. The PSF model estimation error is less than 3% using the results from standard commercial software as an evaluation benchmark. In image deconvolution, we adopted a regularisation-based non-blind deconvolution method using hyper-Laplacian prior. Compared with the widely used methods, the MTF has been improved more than double in high-frequency part. Its processing efficiency and image quality can meet the requirements of RS image products. These results can effectively prove the engineering application value of our proposed method.
查看更多>>摘要：Semantic similarity computing is an important task in NLP, which aims to calculate the semantic similarity between texts. In order to solve the problem of low accuracy of existing calculation methods in Chinese financial intelligent customer services, we propose a text similarity calculation method combining Capsule-BiLSTM network and Chinese Part of Speech (POS) correction. In the text pre-processing stage, POS correction is performed on financial professional words and ambiguous words in the data set to reduce the influence of Chinese word segmentation errors on similarity judgments. Capsule network and BiLSTM were used to obtain local and global information, respectively, the two similarity matrices obtained are fused to determine the similarity. Experimental results show that the proposed method compared with other conventional methods has high precision and F1-value in ATEC data set.
查看更多>>摘要：Emergency Location Routing Problem (ELRP) is a strategic issue in healthcare systems. In this paper, a two-objective mathematical model is presented for the ELRP. The first objective is to maximise the provided services to demand nodes, assuming proper emergency facilities' operation, while the second objective is to maximise the reliability of the facility system to respond the patient's demands even if some facilities fail to operate. To achieve this purpose, a backup system with hybrid ambulance-helicopter transportation system is developed for the situations where the primary system cannot properly serve the patients. The failure probability of the emergency facilities is analysed via robust optimisation. To solve the NP-hard ELRP problem, a balanced exploration-exploitation metaheuristic algorithm based on Genetic Algorithm (GA) and Simulated Annealing (SA), named GASA, is proposed. Research findings by comparison of the GASA simulation results with a commercial solver demonstrate the higher efficiency of the proposed method.
查看更多>>摘要：The Internet of Things (IoT) is an innovative and revolutionary technology. It promises to serve humanity and improve our lives by integrating a huge number of heterogeneous end devices and providing diverse interesting services. However, security remains the main key for the success of any technology. In this paper, we aim to provide structural guidelines to secure IoT-based smart cities. We should consider the IoT-based functional requirements in parallel with the security requirements based on the security triad: Confidentiality, Integrity and Availability. Moreover, we should consider the security challenges of smart cities based on the cube: Big data, Resource limitation, Interoperability, and Scalability. We then present solutions to the main threats. Finally, we assess vulnerabilities of some services offered in an IoT-based smart cities context. We show that current available services still suffer from several vulnerabilities and need associated security countermeasures for the success of the IoT-based smart cities.
查看更多>>摘要：Terrorist attacks, chemical attacks, rapes and other socially sensitive incidents are shared on social media to gain attention from the world. Microblogging sites like Twitter become flooded with these root events and their sub events as they evolve over time, such events are referred to as cross events. Cross event detection is critical in determining the nature of events. The event detection is based on tweet segmentation using the Wikipedia title database. Segment clustering is done based on a similarity measure by encoding the tweets in the form of vectors using the BERT Model. We proposed the Cross Event Evolution Detection framework, which detects cross events that are similar in their temporal nature and result from main events. The experimental results on a real Twitter data collection shows the effectiveness of our proposed framework for both cross event detection and topic evolution algorithm during the evolution of topics and cross events.
查看更多>>摘要：Attribute Partial Order Structure Diagram (APOSD), as an emergent model of concept cognitive learning, has gathered lots of attention due to its good ability of visual knowledge representation. However, there are few in-depth studies on the theory and method of rule acquisition based on APOSD. This paper explores the issue shown above. Firstly, we propose two types of rules, i.e., loose-type rules and rigorous-type rules. Secondly, we investigate the acquisition methods of the above two types of rules. Extraction of non-redundant loose-type rules is discussed as well. Thirdly, we study the method of using the extracted rules for classification. Finally, we conduct comparison experiments on four chosen UCI data sets with classical machine learning algorithms. Results indicate that the method presented in this article performs well and has the advantages of visual expression of knowledge structure and semantic interpretation.
NSTL