In order to improve the detection speed and decrease the false alarm rate of the AdaBoost algorithm in high resolution color image, a multi-feature fusion method is proposed. It adopts cascade strategy to combine multi-fuse classifiers. Benefiting from the complementary of multi-feature, the proposed method can achieve high performance. Experimental result demonstrates that the proposed method has better global precision and is less time consuming.
The existed model on information-theoretically Secure Message Transmission(SMT) assumes t
According to the characteristics of the datasets on the sphere, the algorithm of the K-Nearest Neighbor query based on the query axis (PAM) and the algorithm of the K-Nearest Neighbor query based on the query circular planar(PCM) are presented. Theoretical research and experimental results show that both the two methods can handle the problem of the K-Nearest Neighbor query on the sphere, compared with the PCM, PAM has advantages on the memory capacitance and the query efficiency, but the expansibility of PAM is poor and PCM has high scalability.
Aiming at the problem of large test cases and inefficient software test caused by test requirement change in regression test, a regression test case reduction model is proposed. This model analyzes the association between module, test requirement and test case longitudinally, analyzes association horizontally and reducts among test requirements, and then generates and reducts test cases to realize effective regression test for modified software.
The disadvantages of the HF channel is that it causes a low channel utilization ratio in the HF Wireless Token Ring Protocol(WTRP) network, and it decreases the throughput during the token-ring duration. This paper proposes a modified input queuing switches model for IP gateway in this network according to the characters of HF data communication. The theoretical and simulation researches show that this input queuing switches model can solve the problem of channel resource wasting, make the high priority data to be served first and increase the throughput of the whole net.
This paper derives the capacity formula of time-division half-duplex MIMO decoder-forward relay channel, and compares the performance between time-division half-duplex MIMO relay system and general cellular system by using Monte-Carlo simulation. Simulation results show that, due to the use of an appropriate relay selection algorithm, time-division half-duplex MIMO relay system is better than general cellular system in capacity, coverage. It is prominent that relay user need not use a high transmit power to obtain capacity gain.
This paper proposes an improved broadcasting mechanism and a node-maintaining algorithm based on the analysis of PeerCast broadcasting mechanism and node-maintaining algorithm. Experimental results show that the proposed scheme reduces the time of reconnection and improves the success rate of reconnection when disconnected in PeerCast. And it decreases the times of visiting YP(Yellow Page) after disconnection which reduces the stress of YP.
To enhance the coverage of video sensor network, this paper proposes a coverage-enhancing algorithm using Voronoi diagram based on the adjustable direction sensing model. The algorithm replaces the sector of sensing area with centroid point and constructes a Voronoi diagram by using centroid points. In order to eliminate the sensing overlapping regions and coverage holes, it makes uniform distribution of centriod points uniformly by adjusting sensing direction of nodes. Thereby it enhances the performance of the entire sensor network coverage. Simulation experiments show that the proposed algorithm can improve the coverage area and has rapid convergence.
Network performance is restricted by funnel effect of the data collection sensor networks. With the static data collection sensor networks in mind, based on spatial multiplexing and reduceing the number of hop-step principle, this paper designs construction of multi-mode-level network to achieve different levels of data packets streaming, in order to solute the sink and its few neighboring nodes bottleneck problem effectively, reaches the purpose of easing congestion and reducing transmission delay. Experimental results show that the method can reduce the communication jamming between nodes, lighten data transmitting burden of the nodes effectively, and obtains the network performance.
Network attacker always wishes to penetrate into internal networks, direct access to more important host computers and get higher privilege on victim computers. Based on this assumption, some kinds of greedy policies are analyzed in network attack. The model is constructed and attack graph is limited under such greedy policy. Experimental results show that the attack graph is smaller in size and easier to understand than normal method without losing key attack paths.
This paper proposes a kind of adaptive blind watermark algorithm based on lifting wavelet transform. The watermark is encrypted through new arnold transform, and the original image is decomposed through lifting wavelet transform. Three high frequency coefficients of corresponding locations in the same level are sorted, and come into being a section between the maximum and the minimum which is divided into 2Q-1 sub-sections. The eigenvalue is in the different sub-zone, when the information embedded is different. The watermark can be extracted according to the locations of eigenvalue. This is a kind of blind watermark extracting method. Experimental result shows that this algorithm has a good robustness and recognizable character under many kinds of assaults.
An intersection fuzzy signal control method with bus priority is presented. The buses are given a priority spatially by the bus-only lanes and the bus-priority approaches. Meanwhile, the buses are given a priority temporally by the traffic signal control algorithm with bus priority. The core module of the signal control algorithm consists of green phase module, red phase module and decision module, of which the fuzzy control strategies are designed respectively. The control target is to make bus-weighted vehicle average delay minimum. Simulation results show, compared with fixed-time control, this control strategy not only can reduce bus mean delay significantly, but also can improve social vehicles’ transit efficiency. So this control strategy can be put into practice.
While single service does not match the users’ requirements, automatic service composition composes several services to accommodate the users by using the semantic description of service. In multi-ontology environment, semantic heterogeneity becomes a new challenge for automatic service composition. Aiming at this problem, this paper proposes a service composition architecture based on Ontology Community(OC). The architecture converts service composition across ontologies into that in one OC with the help of ontology translation. Therefore, semantic heterogeneity is avoided. Based on this, it proposes a novel service composition algorithm using classic AI plan.
This paper proposes two approximation algorithms for LCS problem of three input strings. Even if the approximation factors of the two algorithms are still 1/|?|, they can both get much better results in comparison of Gotthilf’s approximating LCS algorithm in most cases. Linear approximating LCS algorithm, takes O(n) time and O(n) space, where n is the length of the longest input sequence. It can work well in the case of large scale input strings. Recursive approximating LCS algorithm, whose time and space complexities both are O(nlogn), can receive the best precision among these algorithms at the most time. It is applicable to meet the desire of good precision for corresponding LCS problem. The two algorithms can be used to solve the LCS problem of multiple sequences and CLCS problem of multiple sequences. Experimental results prove the validity of the two algorithms.
Telephone traffic load of monthly busy hour in two states of Xinjiang are predicted by the method of Support Vector Regression(SVR) combining with Differential Evolution strategy(DE-strategy). The hyper-parameter of SVR is optimized via the DE-strategy and the MAPE criteria is defined as the objective function. Telephone traffic load of monthly busy hour is forecasted by the optimized SVR, the predicted result is compared with the method of grid search and RBF neural network. A better prediction result is obtained by the generalization property of SVR combining with searching property of DE-strategy.
The influence of different band strategies are analyzed. A new bandwidth strategy is put forward. Through structuring an ellipse being similar with target window’s inscribes ellipse, it uses major axis and minor axis to identify bandwidth. This bandwidth strategy is integrated into Mean-Shift target tracking algorithm. Experimental results show this strategy has good effect on tracking target and reduces the time complexity.
In FCMBP algorithm, there is no effective generation algorithm in the database of the translation equivalence classes of high-order fuzzy equivalent standard matrix in the past time, and it is very complicated to define the corresponding database of similar parameters system for each translation equivalence class of fuzzy equivalent standard matrix. In order to solve the problems, this paper proposes an automatic generation algorithm for the database of the translation equivalence classes with the low to high standard fuzzy equivalent matrix, and an algorithm for generating the corresponding database of the equivalent classes of similar parameters system. The case proves that the algorithm effectively solve the problem of automatic generation for the database of the translation equivalence classes of high-order fuzzy equivalent standard matrix.
By taking advantages of the characteristic of Gray code that can extend or remain the block character of binary images, this paper proposes an improved Non-symmetry and Anti-packing Model with K-lines(NAMK) representation algorithm for color images, which is called NAMKG algorithm. The description of NAMKG algorithm is presented and the storage structure and the total data amount of the algorithm are analyzed. Theoretical and experimental results show that, compared with NAMK and Linear Quadtree(LQT), NAMKG can effectively reduce the number of subpatterns and reduce the storage room.
To raise the resolution of the image, this paper presents an improved Adaptive Morphologic Motion Compensation(AMMC) de-interlacing algorithm. It is based on reversibility of wavelet transform to employ the wavelet transform for the post-processing stage. It preserves the edge, and reduces the interpolation error of the motion compensation by using the spatial-temporal correlation and weighting the wavelet coefficients. Experimental results show that the algorithm can reduce the ambiguity of the image edge, and accomplish the better image quality.
Aiming at the detection of infrared faint blob-shaped moving target under complicated background, this paper presents an image clutter suppression method based on wavelet domain Hidden Markov Tree(HMT). A wavelet-domain HMT model is used to accurately capture the dependencies across low frequency scales. It uses Bayesian criterion to estimate image background wavelet coefficients, refers to clutter suppression model to get clutter suppression image signal noise adding model. Gaussianity and independency of residual noise are also verified by using Kendall rank correlation coefficient and Friedman statistic.
Traditional image retrieval system has the semantic gap between image’s labeling and visual characteristic. These systems do not consider the image labeling’s latent semantic relations, and consider the personalization little. In view of these questions, this paper proposes a new personalization image recommendation method. It uses the image’s labeling semantics and vision characteristic to analyze the image, and uses the hyperbolic model to display the image. Through the multianalysis of the semantics relation in image and the original data of image, a prototype system is implemented, and this model is proved feasibility and validity.
This paper researches twenty-four space forms of three-dimensional Hilbert curve primitives, and proposes a n-order three-dimensional Hilbert curve generation algorithm based on the orientation of the primitives. It puts the pixels into a space cube array, sequentially scans them and stores them based on the traversal sequence of three-dimensional Hilbert to realize image scrambling. Experimental results show that three-dimensional Hilbert curve scrambling has a stronger image scrambling degree and higher encryption security compared with the two-dimensional Hilbert curve.
The promotion of efficiency of transporting system is significant to the Semiconductor Wafer Fabrication System(SWFS). To reduce the average traveling time, the average waiting time, and blockages in the Interbay system, this paper proposes a new real-time method using Hungarian algorithm. Simulation results based on data of 300 mm wafer fabrication systems show that this method outperforms the traditional schedule approaches in term of overall efficiency and performance of Interbay material handling system.
WSRF specification is introduced. The WSRF-based architecture for remote experiments is designed by combining the features of FPGA experiments and the state management of WSRF protocol. The resource services of FPGA experiments are designed and implemented by using WSRF.NET, and the correctness of the service is verified by experiments. The result of verification experiments shows that the resource services of FPGA experiments, which are designed by using the WSRF specification, solve the middle state issue of the experiments, and free the time and space of experiments.
BIOS Rootkit is difficult to detect. Aiming at the problem, this paper proposes a behavior characteristics-based BIOS Rootkits detection method. By studying the BIOS Rootkits’ principles and key technologies, this method generalizes, defines and formally describes the behavior characteristics of BIOS Rootkit. Behaviors information is got in decompile phases, and the malicious level is judged according to the rate of a complete BIOS Rootkit formed by those behaviors. Experimental results prove that the method is effective to detect BIOS Rootkits.
The transplantion of operating system for the wireless sensor networks is to modify the hardware-irrelevant layer which makes it work on the different hardware platforms. On this issue, based on the analysis of the structure and hardware-abstract hierarchical on TinyOS 2.0, this paper expatiates how to implement the transplantion on the SoC-based and low-power CC2430 platform and tests the function of the modules. Test results show that the transplanted module works correctly, which lays a solid foundation for the future work.
Agent-oriented scene simulation developing technology based on XML is proposed. Behaviors of simulation entities are described in an Agent-oriented approach, which are stored in an XML file. In this way, the description and control of simulation entities are separated from the simulation software. By applying this method to the deep submergence rescue and life-saving scene simulation, the efficiency of developing a scene simulation system is improved.
This paper represents the background and the significance of video abstraction technology, as well as the three types of research method for video abbreviation. A research method is proposed for video abstraction based on local-space model. In the model, the key frame is got by the analysis of the intra-frame and inter-frame video sequence information, and these key frames are extracted to construct the video segment by double-interpolation. With the DES in the DirectShow software, the video segments are edited and fused into a final dynamic video abstraction. Experimental results show that the video abstraction is generated automatically without any article intervention, which contains the main information of the video but little redundant information.
Aiming at the conflict between the RAM access and the design performance, in the four layer C-to-VHDL reconfigurable compiling framework which is based on Low Level Virtual Machine(LLVM) framework, this paper proposes an optimization algorithm for RAM access. By analyzing the load and store instruction in LLVM’s IR and the data dependence, it creats a dedicated data path to optimize the process of reading and writing to RAM. Experimental results show that the optimized algorithm can effectively reduce the access number to RAM.