Wang Y.,Henan Vocational and Technical Institute
Metallurgical and Mining Industry | Year: 2015
Image feature classification is one of the basic questions of image processing and computer vision and it is also a key step of image analysis. BP neural network has been extensively applied in feature classification and it can classify specific objects or features through early learning; however, BP algorithm also has many defects, including slow convergence speed and easiness to be trapped in local optimum. This paper proposes an image feature classification method based on particle swarm optimization (PSO). It takes the gray image with specific object as the object to be segmented, studies the samples with PSO neural network and gets the training network. Then it takes the pixel matrix of the image as the input vector and puts in the well-trained network for classification. Finally, it realizes the segmentation. The experiment shows that the method of this paper is a feasible one and it has higher convergence speed and stronger robustness. Through the highly-efficient processing, this method can obtain important information and achieve excellent effect when used in the segmentation of the objects in complicated scenes.
Lou S.,Henan Vocational and Technical Institute |
Zhang H.,Zhengzhou University
International Journal of Earth Sciences and Engineering | Year: 2014
Wireless sensor network (WSN) has been widely used in soil monitoring. The localization algorithm is a hot research topic in wireless sensor network. From different perspective, many scholars have put forward some localization algorithm about ranging and non-ranging, when ranging, ranging algorithm causes easily ranging error by environment factor influence, thus damages localization accuracy; The defect of non-ranging algorithm is need to collect center nodes of entire network information in order to estimate position, the communication overhead is large. Aiming at the shortcomings of these localization algorithms, in this paper, proposes a semi-central location algorithm based on support vector regression, the simulation experiments show that the algorithm can relieve the disadvantage of centralized non ranging algorithm which communication overhead is large and the defect of ranging algorithm which is the influence of error accumulation. © 2014 CAFET-INNOVA TECHNICAL SOCIETY. All rights reserved.
Tang H.,Henan Vocational and Technical Institute
Metallurgical and Mining Industry | Year: 2015
Artificial neural network and the intelligent optimization algorithm are hotpots and the cutting-edge ones of the current information science technology, which are of great theoretical and application significance for the fields of pattern classification and identification as well as prediction, etc. This paper put forward a learning method, which is EACONET (Elitist Ant Colony Optimization NET), to optimize the feedforward neural network based on the elitist strategy ant colony optimization, targeting at the weight optimization of the feedforward neural network, in order to solve problems like prematurity and slow rate of convergence of the ant colony optimization in training the neural network. This method combines the global parallel search with the local certainty of BP network, to search for optimal points. At the same time, it can enhance the rate of convergence, avoid the local extremum, can be applied to the functional approximation and the nonlinear system identification, etc. The results of simulation experiment show the effectiveness of the proposed algorithm.
Pan X.,Henan Vocational and Technical Institute
Naukovyi Visnyk Natsionalnoho Hirnychoho Universytetu | Year: 2016
Purpose: Swarm intelligence is the intelligent behaviour represented by a kind of individuals with no or simple intelligence through any form of cluster and collaboration. The research consems the dual-population self-adaptive hybrid algorithm based on genetic algorithm (GA) and artificial bee colony (ABC). We have obtained some important performance measures, which are helpful for swarm intelligence algorithms. Methodology: We proposed a genetic-bee colony dual-population self-adaptive hybrid algorithm based on information entropy, which uses dual-population structure and independent evolution and which conducts information exchange through information entropy to maintain population diversity and accelerate the evolution process between the two populations when appropriate. Findings: We first analysed the basic structure and characteristics of GA and ABC, and then the dual-population based on GA and ABC, which joined the information entropy, was presented, in the parallel operation of two relatively independent populations to accelerate the emergence of a new individual by competition between the populations; it lias better effects in complex function optimization problems. Originality: We made a combinational study of GA and ABC. Although the current biological intelligent evolutionary algorithm has greatly improved its convergence speed, it is not ideal when optimizing complicated functions. This aspect of research is still relatively few at present. Practical value: We researched the optimization algorithm, which is applied to various research fields. Nowadays, it is a development trend to improve the original algorithm by integrating the intelligent algorithm. Dual-population algorithm can overcome the shortage of separate algorithm, and become more suitable for complex optimization problems. We provided the foundation to search for complex distributed problems without centralized control or global model. © Xiaomeng Pan, 2016.
Li J.-R.,Henan Vocational and Technical Institute |
Zhang S.-Q.,Henan Vocational and Technical Institute
Communications in Computer and Information Science | Year: 2012
Message-Digest Algorithm 5, a hash function which is widely used in the field of computer security, can provide the message integrity protection. But to the field following certain rules, for example, by date of birth as a password, using the MD5 encryption algorithm, can still quickly decipher by MD5 decoding procedures of list form. In view of this situation, on the basis of MD5, putting forward a improved algorithm which has greatly security in data safety, and the improved algorithm is applied to the online shopping system based on ajax framework, for digital signature function of users on the system provides powerful safety security. © 2012 Springer-Verlag.