Video copy detection is a crucial technique for copyright protection. However, the main disadvantages of most existing approaches are high computational cost and low robustness. In this paper, we consider videos as a set of shots and propose a video copy detection framework that extracts video shots’ overall features and spatiotemporal features. To effectively enhance the accuracy of final results, a coarse-to-precise filtration approach is proposed in this paper. In the coarse stage, the video copy shot retrieval is preformed by extracting the features of a video shot based on spatial-chromatic histograms. In the refined stage, the spatiotemporal features improved by quantization encoding are applied to the final verification. The combination of FLANN and “as early as possible to stop” process is adopted to accelerate the detection process in the coarse stage. The experimental results show that the proposed approach is effective in detecting video copies with promising precision and recall rate.
This paper puts forwards a method of using MGabor filter-banks to extract Texture-Color features from digital images, and then to construct a group of support vector machines (SVM) classifiers to automatically and accurately classify color digital images. Successful experiments are conducted on the Simplicity and Brodatz image set and our own Ancient shards image sets. The experiments results show the proposed method can integrate the texture features and color information to further improve distinguishing ability of each category images.
FLcom is a fuzzy propositional logic with contradictory negation, opposite negation and medium negation which corresponds to fuzzy set FScom. Based on theories of FLcom, this paper took a financial decision making for example, adopted Distance Ratio Function to give an evaluation to fuzzy propositional, gave a method to ascertain value of ( in the semantic model of FLcom, gave significance of fuzzy production rule based on FLcom and discussed the application of FLcom in fuzzy decision making.
An air quality forecast model based on the resource allocation network has been established in consideration of the time-varying characteristics of the urban air quality and the effects of a variety of nonlinear factors to the prediction accuracy. We have used the distance criteria and error criteria to allocate hidden layer nodes dynamically or adjust network parameters. In this way, we have got the minimum neural network structure to meet the error requirements and avoid solving the problem of selecting the initial neural network structure and parameters.
Intelligent video surveillance systems can be applied to a wide range of potential applications. In this paper, we propose a new background modeling scheme that draws from the principles of low rank representation. We assume that the underlying background images are linearly correlated. Thus, the matrix composed of vectorized video frames can be approximated by a low-rank background matrix plus the sparse foreground components. Low rank representation can be exactly recovered via convex optimization that minimizes a combination of the nuclear norm and the l1-norm, and this non-convex problem can be solved very efficiently in the inexact Augmented Lagrange Multiplier method. We tested our algorithm on real video, and our approach obtained good results, comparable to the Gaussian Mixture Model method.
In order to maximize energy capturing in wind turbines, the wind turbine generator needs to tune the speed of the wind turbine according to the wind. As it is obvious from the literatures, conventional controllers have low efficiency to obtain the better dynamic quality and robustness based on the mathematic models due to uncertainty of wind speed and wind direction. In this paper, an adaptive fuzzy controller is proposed for the variable speed wind power system with uncertain model parameters and force disturbances. The control scheme is based on indirect adaptive fuzzy method using feedback linearization. The uncertainties are approximated by a fuzzy approximator and an adaptive law improves the approximation. The stability of the proposed controller is proved using lyapunov method. Simulation results indicate that the strong robustness and better tracking performance can be achieved rather than a conventional PI controller.
Under the cloud computing environment, the privacy-preserving fuzzy comprehensive evaluation problem is proposed and solved, which has not been studied in the area of secure multi-party computation. Besides, through the analysis of the protocol's security, an improved protocol whose security is higher is given based on blind-and-permute protocol and privacy-preserving protocol of finding the maximum component's position, whose security is stated using the simulation paradigm at last. The experimental results map out the correct possibilities for our algorithm.
CAPTCHA is a simple test that is designed to be easily generated by computers and easily recognized by humams, but difficult for computers to solve. It is now almost a standard security technology. The most widely deployed CAPTCHAs are text-based schemes, but to CAPTCHAs, segmenting the connected and distored characters is still an unsolving problem. In this paper, we proposed a Community Divided Model algorithm which based on complex networks to segment these CAPTCHAs. To evaluate the effectiveness of the proposed segmentation algorithm, we conducted several experiments on database which collected some CAPTCHAs from the Internet randomly. The results showed that the proposed algorithm is effective to segment two or more connected and distored characters.
Considering the serial strategy generally used in feature fusion easily leads to curse of dimensionality and two-dimensional matrix for image representation outperforms one-dimensional vector, a novel strategy of parallel complex-matrix-based horizontal and vertical discriminant analysis is developed in this paper. It first respectively utilizes two different images of a subject as the real and imaginary part of a complex matrix, two-step discriminant analysis, namely horizontal LDA and vertical PCA, is then performed in the complex feature space. The experimental results demonstrate that the proposed method is more promising and effective.
Considering the issue of decentralized collaborative target tracking architecture in the sea battlefield for the wide perception and complex sensor networks, firstly a new target calculation mechanism of the collaborative target tracking is proposed. To increase the performance of robustness, self-organization and dynamic adaptability for the information dissemination and sharing strategy, research methods and technical route are discussed in detail on the basis of complex network theory. In order to effectively deal with different kinds of information sources in the sensor networks, a generalized fusion machine is presented by way of DSmT model. The proposed architecture is applicable to the further research of collaborative target tracking technologies in the sea battlefield.
The adaptive beamforming algorithm can aim at the direction-of-arrival of jammer automatically. But when the antenna platform vibrates or interference moves quickly, it is possible that the mismatching occurs between adaptive weight and data due to the perturbation of the interference location. To solve these problems above, a novel algorithm for null widening is presented in this paper. By using quadratic constraint on the interference integration matrix, the proposed approach can control the depth and width of null with different user parameters, so the interferences can be suppressed effectively in such case and the robustness of this adaptive beamforming algorithm is improved. The formulas of this algorithm are deduced and the ranges of Lagrange multiplier value are computed. Simulation results show the effectiveness of the proposed algorithm.
Nowadays, almost every search engine use "Discrete" models, such as boolean logic model. it processes user queries and documents, in the way that will break key words into discrete, unrelated words. It will lose the semantic relation between words, and bring noise in search results. This paper presents a conceptual analysis method, replacing the interrogative word in the query with focus word that is extracted from the query, in order to construct a concept map expressing the connotation semantic information of the query. From the view of Chinese conceptual connotation, we analyze user demands, and restore their retrieval intention, in order to guide the next search, to improve the accuracy of the retrieval system.
Coordination takes the role of integrating a set of individual robots into a whole multi-robot system to accompany tasks. During the past two decades, lots of achievements for the coordination of multi-robot systems were made. In this paper, these results were reviewed from two aspects. The first is from the point of view of that coordinated strategies were generated automatically by mathematical approaches, and the second is from the point of view that coordinated strategies were designed by control engineers. The approaches for generating and describing coordinated strategies are summarized, respectively. The potential future work was discussed especially for the case the goal of exploration is modeling an unknown indoor environment. It was pointed out that a few of coordinated behaviors can be realized by both of quantitative and qualitative approaches.
In recent years, the spatial resolution of a remote sensing image becomes much higher than ten years ago. The research of image processing and analyzing based on traditional low resolution image has already not satisfied the need for getting more accurate information. Identifying particular objects from remote sensing image become more important to Digital City and real-time monitoring. The paper proposes a novel semantic manifold interpretation method of high-resolution sensor image, which uses semantics associated with ground object images to improve object recognition works. Our approach first learns the multiple semantic classes by using a semi-supervised manifold learning algorithm to produce a "semantic manifold" of the ground object, and then the RF(Relevance Feedback) iteration based on manifold ranking algorithm is then run on the semantic manifold. The methods are applied to several high-resolution example images, and some buildings as test objects in images are recognized. Those examples illuminate that the method proposed in this paper is effective and accurate, especially for multi-view, multi-spectral, all-weather remote images.
The forecasting system for medium to long term fishery resources is based on historical production data of specified fish types and those marine environmental factors. As these systems give a macro level prediction of fishery resources in the coming years they provide indispensable references for the planning and management of catching seasons. This paper introduces a new model for the prediction using Windows XP platform and Visual Studio 2010 development environment with C# programming language. Combining correlation analysis and BP neural network, the new model analyzes marine environmental data and fishery historical production data to forecast fisheries in medium to long terms. Experiments applying this model to forecast the squid production in the Pacific Northwest result in an average relative error of about 13.5% as compared with 23.2% error using linear regression analysis. This result proves that the new model has the potential to provide better forecasts for fisheries.
According to the actual power system transmission line, the voltage and current transient simulation models are set up. Four different types of transient signals are obtained and they are voltage oscillation, current oscillation, voltage pulse and current pulse. The wavelet and the lifting wavelet are respectively used to detect the above four kinds of transient signals. The simulation results show that both the wavelet and the lifting wavelet can accurately detect the singularity of transient signal. However, the reconstruction error of the lifting wavelet is much less than that of the wavelet. The algorithm of the lifting wavelet is more accurate than that of the wavelet. Therefore, the lifting wavelet is more suitable for the accurate and real-time operating requirements in the power system.
Izhikevich neuron is a relatively new neuronal model, which has found extensive applications in modeling neuron due to its strong biological plausibility and computational effectiveness. In this work we use the information theoretic method to measure the ability of information transmission of this neuron model. We find that Izhikevich neuron shows low sensitivity to high frequency of random noise; and appropriate noise level can help information transmission through the neuron.
In general, wireless sensor network works by a small battery-powered, or limited energy. Once the wireless sensor network is deployed, the energy of small sensor nodes can not be replaced. So, to improve energy efficiency and extend the survival time of the whole network is a crucial issue. This paper presents a new type of energy balancing algorithm for wireless sensor networks. The BCDCP-M algorithm draws on the main idea of the BCDCP routing protocol. When the base station divides the network, the new algorithm makes the number of cluster heads equal to the optimal number of cluster heads as far as possible. On cluster head election, not only the average energy of the sensor network, but also the remaining energy of the individual node must be taken into account. In data transmission, we use the multi-hop method to select the optimal path. The simulation results show that the survival time of the network in the new BCDCP-M algorithm is 19% longer than BCDCP.
In this paper we propose an automatic salient object extraction method for nature scene. The proposed method first utilizes an algorithm based on visual attention model to obtain a prior knowledge for Graph Cut, and then constructs the weighted graph of Graph Cut based on super-pixels pre-segmented by the improved watershed algorithm in order to accelerate the speed of proposed method. In this framework, Visual saliency map is obtained using chrominance and intensity features in HSV color space, which provides the approximate region that contains salient object to be segmented. Then the salient object region after extension is cropped as input image, and pre-segmented by the improved watershed algorithm into several regions to construct weighted graph. Finally the salient object is obtained by Graph Cut algorithm. Experiment results show that our algorithm can automatically get salient object without human interactions, and speed up the segmentation without decreasing segmentation accuracy.
A person-computer ensemble system is one of time concerned cooperative systems, which performs the secondo in an ensemble played by a computer-controlled piano cooperating with the primo played by a person musician. For expressive performances, it is necessary to prepare very expressive secondo data before the actual cooperative performance, since the system is modified its expression in realtime using the agogic information calculated from the input stream of the primo. In this paper, a generating method of more expressive performance of the secondo, using the notions musical structures and structural functions, are introduced.
