Search

Footprints can provide strong evidence for the detection of criminal cases, and the similarity retrieval of footprint images is generally carried out using hand-extracted image features, which have problems such as low retrieval matching accuracy and slow retrieval speed. In order to solve the above problems, a footprint image retrieval method based on deep supervised binary hash(DSBH) is proposed, and the feature extraction of footprint image is carried out by using the convolutional neural network, which can be combined with the deep hash algorithm to solve the retrieval problem of footprint image. The experimental results can reach 0.980, which proves the effectiveness of this method.

With the increase of gray scale and Flat Panel Display (FPD) size, Subspace Bitwise Scanning strategy with high scanning efficiency can replace traditional scanning method to enhance frame frequency. However, as gray scale increasing, the Direct Searching Strategy(DSS) becomes unfeasible to obtain corresponding scanning matrix of the optimization strategy. Therefore, genetic algorithm introduced accelerates searching FPD weights scanning matrix utilizing its parallelism. Finally a scanning matrix of 128 gray scales whose scanning efficiency reaches 99.99% is found out successfully in about 25 hours.

A new type of inorganic flocculant, poly silicon boron ferric zinc sulfate (PSBFZ), was synthesized from sodium silicate, ferric sulfate, zinc sulfate and sodium tetra borate by co-polymerization. Comparative flocculation behaviors among poly-aluminium chloride (PAC), poly-ferric sulfate (PFS), poly silicate ferric (PSF) and PSBFZ were explored. The influences of pH, temperature, storage time and dosage on flocculation were studied. The results demonstrate that PSBFZ exhibits the best flocculation performance. Under the optimum conditions, the removal rate of turbidity and chromaticity were up to 99.6%, 99.4%, respectively.

In the process of disease diagnosis, determining the types of disease is very important. With the development of DNA microarray technology, the research on huge gene expression profile has become the focus of disease classification. This paper presents a method of classification of gene expression profile based on the adjacency matrix spectral decomposition. First, samples are mapped to a high-dimensional space of points to construct an adjacency matrix, and we can obtain eigenvectors describing the feature of the samples by decomposing the matrix. Finally, use eigenvectors as inputs of the SVM (Support Vector Machine) and KNN (K nearest neighbor) classifiers to classify gene expression profile. In this way sample information can be completely preserved, which enables an approach to making gene expression profile from one without structural information to one with structural information. The validity of this method is verified by comparative experiments.

The power generated by the PV generator depends on the irradiance, temperature, and surface conditions. The PV generator has a nonlinear characteristic, and the power has a maximum power point or multi-local maximum power points under different conditions. Various MPPT methods have been reported, in order to confirm the effectiveness of each control method, experimental or simulation analysis is used, from economical point of view, simulation analysis is more effective. In this paper, a versatile, accurate Matlab/Simulink PV model suitable for use by power electronics specialists has been developed, and has been designed for easy implementation on Matlab/Simulink platform; conventional MPPT methods and MPPT methods suitable for partial shading conditions can be simulated by using this PV model.

In e-commerce, the consultation based on multi-agent has become an active field. We give a multi-agent bilateral multi-project negotiation model which is based on multi-agent blackboard model and uses adaptive genetic algorithm. Firstly, we should set up multi-agent blackboard framework; secondly, we should implement negotiation model using adaptive genetic algorithm; finally, using web technology and jade, we can simulate the model and prove experiment result. The model resolves the problem that there is only one topic in negotiation in e-commerce and .improves transaction significantly.

According to the characteristic of wavelet packet transformation and chaotic neural networks, this paper proposed a new method of discriminating fault line for single-phase ground in distribution system based on the wavelet packet transformation and chaotic neural networks. The wavelet packet chaotic neural networks trained by the post fault zero sequence current and zero sequence voltage are used to discriminate the fault line. This method is suitable for any distribution system, not influenced by load problem of the traditional methods that select fault line incorrectly. The EMTP shows that the proposed method is accurate and reliable, and has great application potential.

In e-commerce, the negotiation based on Multi-Agent has become an active field. This paper brings a new Multi-Agent Auto Negotiation Model which is done using multi-Agent System, fuzzy evaluation algorithm and game theory. Firstly, we design negotiation model; secondly, we analyze the flow; finally, we build a negotiation platform in according to combination with Web technique and jade. The model solves information asymmetric, achieves auto-bargaining, and guarantees business security.

Blank holder force(BHF) forecasting is an important research direction in sheet metal forming. Support vector machine(SVM) is a novel learning machine based on the principle of structural risk minimization, which can solve the shortcoming of artificial neural networks, such as local optimization solution, lack generalization. In the study, SVM is presented to blank holder force forecasting. Cupshell which is a typical drawing workpiece is taken as the research object. The experimental results show that the proposed SVM provides better prediction capability for blank holder force forecasting than BP neural network. Therefore, SVM is considered as a promising alternative method for blank holder force forecasting.

Fuel cell (FC) systems have been showing up as a promising alternative due their high efficiency, low aggression to the environment, excellent dynamic response, and superior reliability and durability. This paper introduces the electrochemical model of FC and establishes the Simulink model based on it. The results of the model are used to predict the output voltage, efficiency, and power of FC as a function of the actual load current and of the constructive and operational parameters of the cells. Detailed simulation researches on polarization curves have been carried through for FC performances boosting. Parameter influences of FC are mainly discussed on the dynamical performance of PEMFC and some parameters which might be despised but very important are pointed out, especially. Corresponding dynamic behaviors of a typical 500-W BCS stack are presented. In the end, control strategies are designed for FC operation optimization, and some accidental cases are pointed out.

A new control method of induction motor is proposed in this paper to reduce the pulsation of motor torque and magnetic flux linkage, which, as a great disadvantage, exists in the direct torque control system. According to the conclusion, the required voltage vector is provided which can be composed by space voltage vector pulse width modulation. Simulation results show that it effectively solves the problem of torque ripple and flux ripple

Direct torque control (DTC) schemes can produce quick and robust responses, but heir major problem is the high torque and flux ripple, especially when operation conditions result in low speed. This paper proposes a new control strategy. On the basis of conventional direct torque control, a flux and torque ripple minimum controller is introduced in the strategy in order to decrease flux and torque ripple. Simulation results justify the validity of the proposed method to reduce the motor torque and flux ripple