Vol 51, No 4 (2017)

The problem of production rules extraction has been solved based on parallel computing and computational intelligence. The research object is the process of production rules extraction. The purpose of the work is the creation of the production rules extraction method, based on a parallel principle of the construction of the intelligent models, which bring together given data samples in the form of the models based on the decision trees, association rules and negative selection. The developed method allow to significantly reduce the time required for the models synthesis when solving the complex practical problems, characterized by a large amount of the diagnostic data; and the problems, where there is a need to modify the existing diagnostic and recognition models due to the appearance of new information, which is the result of the permanent observation after the state of the research objects and processes. At the same time, the capability of the synthesis of the models that have the high approximating and generalizing abilities is provided.

This study made an inductive discussion on the definition and characteristics of computer viruses and analyzed virus transmission models as well as virus control. Like tumor cells in human body, computer viruses will rapidly transfer and disperse if they are not controlled, indicating a poor stability. To investigate the transmission control of computer viruses, a dynamical model was established for virus transmission and the concept of “equivalent day” was introduced to analyze the dynamic characteristics of discrete transfer of the model and the stability of virus-free equilibrium points and endemic equilibrium points. The characteristic value was obtained by calculating the equation of the model. Moreover, the necessary and sufficient conditions for virus-free and endemic equilibrium points of the model were obtained by proof using Lyapunov first method and disc theorem. Then the model was compared with SIS and SIR models. Finally, a control item was added based on the virus discrete transmission model and the transformation trends of the number of the infected principle machine and the infective principal machine before and after the addition of the control item were compared. In this way, the optimal control strategy for virus transmission model was designed and the effectiveness of the optimal control was verified.

Superficial and electroplating defects are the most commonly seen flaws on the lens collar. The former arose from cutter offset or chip winding during the cutting process, while the later occurred if the surface was stained with rough or foreign material during the electroplating process. Relying on human inspection to ensure quality of a lens collar was time consuming and accounted for occupational injury. Thus, implementation of automatic inspection technology became invertible in the mass production environment. Since the texture on the surface of lens collar was not only regular but also statistical, the system used image restoration based on discrete Fourier transformation (DFT) to detect defects embedding on those two types of textures.

The efficiency of image enhancement algorithms depends on the quality and processing speed of image enhancement. There are many algorithms to implement image enhancement using wavelet theory. These algorithms have one thing in common: they all capture image details by decomposing low frequency sub-images. In fact, a lot of details in high-frequency sub-images are also found. Enlightened by the above-mentioned facts, a novel medical image enhancement method based on wavelet decomposition is proposed by adding details from the high-frequency sub-images and decomposing the image specially with ant-symmetric biorthogonal wavelet instead of some traditional wavelets. It not only improves the image enhancement, but also overcomes the shortcomings of large computation with faster computational speed and satisfies the real-time requirement in edge detection. Simulation experiments of mammographic images are implemented by Matlab with several different methods, the results show that the proposed method is superior to some popular methods, such as histogram equalization and wavelet nonlinear enhancement.

This article proposes special digital filtering structures formed by one multiplier, one adder and a delay line. Such structures can be used to synthesize finite impulse response Hilbert filter. The advantage of such approach is reduced number of multipliers, compared to standard implementation. The quotients of structures can be calculated by local optimization methods. This article shows an example of the 30-th order Hilbert filter calculation, which requires only 6 multipliers for implementation.