Kangming L.,Huanghuai University
RISTI - Revista Iberica de Sistemas e Tecnologias de Informacao | Year: 2016
The wide application of pedestrian detection method makes it an important part in the computer field, which provides significant support in multiple domains as video monitoring, driver assistance system and intelligent robots, etc. Based on full investigation of related technologies as pedestrian detection and deep learning, this paper summarized the main difficulties of pedestrian detection and problems of the current method, specifically studied the pedestrian detection method based on deep learning, and proposed an improved pedestrian detection method with DBN classification algorithm on account of the problems existed in current shallow classification method as over amount of training samples and weak fitting ability of complicated functions, etc. First, the research improved the DBN input mode by establishing T-distribution-included function outermost layer nodes RBM (Restricted Boltzmann Machine) input terminal, and converted the extracted pedestrian feature information into the classifier recognizable Bernoulli distribution through the outmost layer structure of the input terminal. Then, BP neural network is used to establish the output terminal of the classification structure to realize the revert transmitting of the classification error message, tune the parameters of classification structure and optimize the structure of the classifier continuously. The experiment shows that the improved DBN pedestrian detection method has better performance than that of the classic shallow-layer classification algorithm. The detection speed of the improved algorithm can satisfy the use requirement.
Shao L.-F.,Huanghuai University
Electronic Journal of Geotechnical Engineering | Year: 2015
Rock burst is a geological hazard that frequently occurs in deep underground engineering and thus its prediction is of important significance. The existing rock burst prediction research focuses on the prediction methods, rather than the importance, association cost of consequence and combined relations of indicators. In this paper, a new method called "ordered logistic regression" is used to predict and analyze rock burst. This method adopts six basic indicators are: the maximum tangential stress in surrounding rock mass σθ, the compressive strength σc, tensile strength st, the stress coefficient sθ/σc, rock brittleness coefficient σc/σt and elastic energy index Wet. Ordered logistic regression models for rock burst prediction are built in this paper. The importance of rock burst indicators is selected on the basis of 108 groups of rock burst samples. The results show that the indicator combinations (sθ/σc, Wet; sθ, σc, Wet) are more suitable for establishing an ordered logistic regression model for rock burst prediction. The obtained cumulative probability expressions can be referred to in underground engineering rock burst prediction in future. Analysis of rock burst indicators indicates that the indicators can be ranked by importance degree as sθ > sθ/σc > Wet > σc > σt > σc/σt, and ranked by sensitivity as sθ/σc > Wet > σc/σt.σtand σc/σt are not suitable for rock burst prediction. © 2015 ejge.
Li H.,Huanghuai University
RISTI - Revista Iberica de Sistemas e Tecnologias de Informacao | Year: 2015
Objective: The research results of computer intelligent vision application technology in manufacturing industry are analyzed. Method: Analysis and summary of computer intelligent vision application. Process: Based on the classification of visual inspection, the construction of visual model and the characteristics of computer vision, the relationship between intelligence, vision and computing in computer vision was illustrated. The visual detection technology Respectively is introduced to be used in drawing die hole shaped visual detection system, the surface of the steel plate automatic detection system and image acquisition system. The application status, development trends and difficulties of visual inspection was analysed, and the problems were clarified in the development of computer vision inspection technology in china. Result& Analysis: In recent years, computer vision application technology has been improved, but the problem is very serious. Result: The prospect of computer vision application technology is very bright, and its development will promote the development of manufacturing industry.
Shao L.-F.,Huanghuai University
Electronic Journal of Geotechnical Engineering | Year: 2014
Luanchuan County lies in the mountains of Western Henan Province. With complicated geomorphy and topography, this area is characterized by great gully slope, complex structures, intense human engineering activities and heavy and concentrated rainfall, which have exposed the area to frequent debris flow events. On July 24, 2010, due to a widespread rainstorm, most of the county was attacked by catastrophic debris flow, causing heavy human casualties and property damages. This paper is designed to analyze material sources for the 7.24 debris flow event with a view to providing reference for the post-disaster reconstruction and prevention. Our analysis results indicate that the debris flow sources inside Luanchuan County are concentratedly distributed and fault activities as well as human activities provide abundant material sources for the formation of debris flow; human activities, including cultivation and mining, have an important share of contribution to the increased debris flow sources in the county. On these grounds we also put forward some proposes of how to improve the debris flow source conditions of the county. © 2014 ejge.
Liu W.,Huanghuai University
Optoelectronics and Advanced Materials, Rapid Communications | Year: 2011
We present a detailed analysis of the light absorption in the coaxial nanowire (NW) of Si core and nonabsorbing dielectric shell which has great potential in solar cells. We have calculated the absorption efficiency within the framework of the Lorenz-Mie light scattering theory to investigate the effects of the core radius, the shell thickness, and the shell refractive index on the light absorption. We have found that the photocurrent can be enhanced up to ~135% for the coaxial NW of the core radius of 150 nm by tuning the shell thickness and the shell refractive index, as compared to Si NW.