Search

Optical motion capture (MOCAP) is a commonly used technology to record the motion of non-rigid objects with high accuracy in 3D space. However, the MOCAP data has to be processed further before it can be used. The scattered reconstructed motion data must constitute a human configuration by labelling process according to the predefined template, and the missing markers have to be reconstructed to produce a stable motion trajectory. In this work, we propose a novel labelling method for motion sequences. First, a novel graph matching method is employed to determine the connection relationship of the scattered motion data for a single frame. Then, Kalman filtering is used for tracking in the motion sequence. As for the challenge coming from missing markers, we propose a new motion data preprocessing method considering the bone length constraint, which represents the information of variation in the relative position of adjacent markers. The processed motion data is input into a Long-Short Term Memory (LSTM) model to recover the missing markers and de-noise the motion data. The experiment conducted on our own dataset proves that our labelling method achieves a similar effect to Cortex, which is a commonly used commercial motion data analysis software. The experiment on CMU dataset demonstrates that our missing marker reconstruction method can achieve an art-of-state result. The labelling code will be pulished on https://github.com/Lijianfang6930/Graph-Matching-for-Marker-Labelling

With the continuous innovation of new-generation information technology and its accelerated integration with manufacturing industry, industrial internet platforms (IIPs) are rapidly emerging worldwide. Construction and application of IIP has become a new focus in international competition for leading enterprises, and also a new direction of industrial development for many countries worldwide. However, the development of IIP is still in the stage of exploration, and the industry sector still lacks unified understanding of the IIP. Therefore, this study firstly proposes a reference architecture of IIP to clarify its framework and core functions, so as to provide a general reference model for the industry to understand and jointly promote construction of IIP. Secondly, an assessment system is proposed to evaluate the usage of IIP. The assessment framework is composed from three domains namely the foundation, key capability, value and benefit. Finally, the practical value of the reference architecture and the assessment framework of IIP is verified by an industry practice.

With the large-scale deployment of cloud datacenters, high energy consumption and serious service level agreement (SLA) violations in datacenters have become an increasingly urgent problem to be addressed. Implementing an effective virtual machine (VM) consolidation methods is of great significance to reduce energy consumption and SLA violations. The VM consolidation problem is a well-known NP-hard problem. Meanwhile, efficient VM consolidation should consider multiple factors synthetically, including quality of service, energy consumption, and migration overhead, which is a multi-objective optimization problem. To solve the problem above, we propose a new multi-objective VM consolidation approach based on double thresholds and ant colony system (ACS). The proposed approach leverages double thresholds of CPU utilization to identify the host load status, VM consolidation is triggered when the host is overloaded or underloaded. During consolidation, the approach selects migration VMs and destination hosts simultaneously based on ACS, utilizing diverse selection policies according to the host load status. The extensive experiment is conducted to compare our proposed approach with the state-of-art VM consolidation approaches. The experimental results demonstrate that the proposed approach remarkably reduces energy consumption and optimizes SLA violation rates thus achieving better comprehensive performance.

This research revisits the fundamental problem of processing privacy-preserving Boolean queries over outsourced databases on untrusted public clouds. Much current searchable encryption (SE) schemes try to seek an appropriate trade-off between security and efficiency, yet most of them suffer from an unacceptable query leakage due to their conjunctive/disjunctive terms that are processed individually. We show, however, this trade-off still can be deeply optimized for more security. We consider a Boolean formula as a set of deterministic finite automatons (DFAs) and propose a novel approach to running an encrypted DFA, which can be effectively and efficiently processed by the cloud. We give three constructions for conjunctive, disjunctive, and Boolean queries, respectively. Their notable advantages are single-round, highly-efficient, adaptively-secure, and leakage-minimized. A lot of experiments are made to evaluate overall efficiency. Testing results show that the schemes achieve enhanced security almost without sacrificing anything of search efficiency.

Visual Question Answering (VQA) is a challenging multi-modal learning task since it requires an understanding of both visual and textual modalities simultaneously. Therefore, the approaches used to represent the images and questions in a fine-grained manner play key roles in the performance. In order to obtain the fine-grained image and question representations, we develop a co-attention mechanism using an end-to-end deep network architecture to jointly learn both the image and the question features. Specifically, textual attention implemented by a self-attention model will reduce unrelated information and extract more discriminative features for question-level representations, which is in turn used to guide visual attention. We also note that a lot of finished works use complex models to extract feature representations but neglect to use high-level information summary such as question types in learning. Hence, we introduce the question type in our work by directly concatenating it with the multi-modal joint representation to narrow down the candidate answer space. A new network architecture combining the proposed co-attention mechanism and question type provides a unified model for VQA. The extensive experiments on two public datasets demonstrate the effectiveness of our model as compared with several state-of-the-art approaches.

Searchable encryption provides an effective mechanism that achieves secure search over encrypted data. A popular application model of searchable encryption is that a data owner stores encrypted data to a server and the server can effectively perform keyword-based search over encrypted data according to a query trapdoor submitted by a data user, where the owner’s data and the user’s queries are kept secret in the server. Recently, many searchable encryptions have been proposed to achieve better security and performance, provide secure data updatable feature (dynamics), and search results verifiable capability (verifiability). However, most of the existing works endow the data user an unlimited search capacities and do not consider a data user’s search permissions. In practical application, granting search privileges for data users is a very important measure to enforce data access control. In this paper, we propose an attribute-based searchable encryption scheme by leveraging the ciphertext-policy attribute-based encryption technique. Our scheme allows the data owner to conduct a fine-grained search authorization for a data user. The main idea is that a data owner encrypts an index keyword under a specified access policy, if and only if, a data user’s attributes satisfy the access policy, the data user can perform search over the encrypted index keyword. We provide the detailed correctness analyses, performance analyses, and security proofs for our scheme. The extensive experiments demonstrate that our proposed scheme outperforms the similar work CP-ABKS proposed by Zheng on many aspects.

With the enhancement of people's health awareness, more and more users are willing to wear portable micro-health monitoring equipment and communicate with remote medicine center for real-time diagnosis. Although, under normal circumstances, users' health status can be detected at any time, in extreme circumstances, such as earthquakes, how to make the medical center monitor user data for a long time for rescue will be of great significance. In this paper, we will study the networking of portable wearable devices based on wireless sensor networks. We mainly use minimal connected dominating sets (MCDSs) to organize nodes in extreme environments effectively, form virtual backbone networks, send data to the rescue or medical personnel, and maximize network lifetime. Specifically, we propose an adverse dominator selection procedure (ADSP), where the dominators are selected by their children-independent nodes. The ADSP has two versions, which are Independent node degree-based Adverse Dominator Selection Procedure (IADSP) and residual Energy-based Adverse Dominator Selection Procedure (EADSP). Based on IADSP and EADSP, two approximation MCDS construction algorithms named Independent node degree based MCDS (IMCDS) and Energy-efficient Independent neighbor-based MCDS (EIMCDS) are proposed, respectively. Both of them have the message complexity as O($N\Delta $ ). The performance ratio of IMCDS has an upper bound as O($\sqrt {N}$ ). The simulation results show that IMCDS and EIMCDS perform well in terms of CDS size, and the routing algorithm based on EIMCDS has better energy efficiency performance than that of IMCDS and classical routing protocol.

Clustering, which explores the visualization and distribution of data, has recently been widely studied. Although current clustering algorithms such as DBSCAN, can detect the arbitrary-shape clusters and work well, the parameters involved in these methods are often difficult to determine. Clustering using a fast search of density peaks is a promising technique for solving this problem. However, the current methods suffer from the problem of uneven distribution within local clusters. To solve this problem, we propose a new density peak based clustering algorithm employing a hierarchical strategy, namely, HCFS, which consists mainly of two stages. In the first stage, the HCFS estimates the density and distance of each point. The points with higher density and distance are selected as candidate centers, and then subclusters centered on them are further obtained. In the second stage, considering that adjacent subclusters based on certain candidate centers are highly similar and connected within the same cluster, we propose a new mechanism for measuring dissimilarity and connectivity between the subclusters. Those highly similar and connected subclusters are merged to increase the dissimilarity between different clusters and to obtain the final clustering results. The experiments conducted on a large number of datasets show that our method can effectively identify unevenly distributed clusters and yield better or comparable performance for different datasets.

Software-defined networks (SDNs) are a trend of research in networks. Rule placement, a common SDN operation, becomes a challenging problem due to the capacity limitation of devices in which a large number of rules need to be deployed. Prior works mostly consider rule placement in a single device. However, the position relationships between neighbor devices also have influences on rule placement and should be considered. Our basic idea is to classify the devices position relationships into two categories: the serial relationship and the parallel relationship, and we present novel strategies for rule placement based on the two different position relationships. There are two challenges of implementing our strategies: to check whether a rule is contained by a rule set or not and to check whether a rule can be merged with other rules or not. To handle the challenges, we propose a novel data structure called OPTree to represent the rules, which is convenient to check whether a rule is covered by other rules. We design an insertion algorithm and a search algorithm for OPTree. Extensive experiments show that our approach can effectively reduce the number of rules while ensuring placed rules work. On the other hand, the experimental results also demonstrate that it is necessary to consider the position relationships between neighbor devices when placing rules.

Virtualization technology, as a key technology in cloud computing, makes the virtual machine placement (VMP) play an important role in improving the energy efficiency of data centers. In this paper, an energy-aware algorithm named GATA is proposed for the VMP problem. It combines the genetic algorithm with the tabu search algorithm. The goal is to obtain an optimal VMP scheme to achieve energy efficiency while maximizing load balance among various resources. The algorithm is compared with two meta-heuristic algorithms and a newly proposed algorithm that is based on ant colony algorithm. The execution time of these algorithms is also discussed. The results show that the proposed algorithm is superior to those methods mentioned above.

Over these years, object tracking algorithms combined with correlation filters and convolutional features have achieved excellent performance in accuracy and real-time speed. However, tracking failures in some challenging sequences are caused by the insensitivity of deeper convolutional features to target appearance changes and the unreasonable updating of correlation filters. In this paper, we propose dual model learning combined with multiple feature selection for accurate visual tracking. First, we fuse the handcrafted features with the multi-layer features extracted from the convolutional neural network to construct a correlation filter learning model, which can precisely localize the target. Second, we propose an index named hierarchical peak to sidelobe ratio (HPSR). The fluctuation of HPSR determines the activation of an online classifier learning model to redetect the target. Finally, the target locations predicted by the dual learning models mentioned above are combined to obtain the final target position. With the help of dual learning models, the accuracy and performance of tracking have been greatly improved. The results on the OTB-2013 and OTB-2015 datasets show that the proposed algorithm achieves the highest success rate and precision compared with the 12 state-of-the-art tracking algorithms. The proposed method is better adaptive to various challenges in visual object tracking.

Reconstructing precise dynamic points with multiple camera systems (MCSs) is a pivotal work in many computer vision applications, such as motion capture. However, the deviation of 2-D position leads to frequent mismatch when searching for correspondence from multi-view. This paper puts forward a two-stage framework based on passive optical motion capture system to reconstruct precise dynamic points with MCSs. Our proposed method improves the performance of calibration and matching simultaneously. In the calibration stage, the extrinsic parameters of numerous cameras are calibrated synchronously via an L-shaped frame, where the position of four reference points is optimized with multiple geometric constraints. Bundle adjustment occurs after calibration. In the reconstruction stage, we propose a novel sparse multi-view matching method called cyclical voting, which includes multiple pairs of global voting and in-group voting. Point residual method is proposed to exclude outliers in matching groups further. The experiments show that our proposed method can decrease mismatching significantly and achieve commendable reconstruction results compared with Cortex (one of the most successful commercial motion analysis software).

With the considerable growth of cybersecurity risks in modern automobiles, cybersecurity issues in the in-vehicle network environment have attracted significant attention from security researchers in recent years. Enhancing the cybersecurity ability of in-vehicle networks while considering the computing resource and cost constraints become an urgent issue. To address this problem, a novel information entropy-based method is proposed in this paper, which uses a fixed number of messages as sliding windows. By improving the sliding window strategy and optimizing the decision conditions, the detection accuracy is increased and the false positive rate is reduced. Experimental results demonstrate that the proposed method can provide real-time response to attacks with a considerably improved detection precision for intrusion detection in the in-vehicle network environment.

Information centric networking (ICN) and software-defined networking (SDN) are two emerging networking paradigms that promise to solve different aspects of networking problems. ICN is a clean-slate design for accommodating the ever increasing growth of the Internet traffic by regarding content as the network primitive, adopting in-network caching, and name-based routing, while SDN focuses on agile and flexible network management by decoupling network control logic from data forwarding. ICN and SDN have gained significant research attention separately in the most of the previous work. However, the features of ICN have profound impacts on the design and operation of SDN, such as in-network caching and data-centric security. Conversely, SDN provides a powerful tool for experimenting and deploying ICN architectures and can greatly facilitate ICN functionality and management. In this paper, we point out the necessity of surveying the scattered works on integrating SDN and ICN (SD-ICN) for improving operational networks. Specifically, we analyze the points of SD-ICN strengths and opportunities, and discuss the SDN enablers for deploying ICN architectures. In addition, we review and classify the recent work on improving the network management by SD-ICN and discuss the potential security benefits of SD-ICN. Finally, a number of open issues and future trends are highlighted.

Cybersecurity is increasingly important for the safety and reliability of autonomous vehicles. The controller area network (CAN) is the most widely used in-vehicle network for automotive safety-critical applications. Enhancing the cybersecurity ability of CAN while considering the real-time, schedulability, and cost constraints becomes an urgent issue. To address this problem, a real-time, and schedulability analysis-guaranteed security mechanism [identification hopping CAN (IDH-CAN)] is proposed in this paper, which aims to improve the security performance of CAN under the constraints of automotive real-time applications. In order to support the operation of the IDH-CAN mechanism, an IDH-CAN controller is also designed and implemented on a field-programmable gate array, which can work as a hardware firewall in the data link layer of CAN to isolate cyberattacks from the physical layer. Meanwhile, to maximize the information entropy of the CAN message ID on the physical layer, the ID hopping table generation and optimization algorithms for IDH-CAN are also proposed. Then, information security evaluation experiments based on information entropy comparison are deployed. The simulation and practical evaluations demonstrate the effectiveness of the proposed mechanism in defending reverse engineering, targeted denial of service, and replay attacks without violating real-time and schedulability constraints.