Xiaofeng Lian

Research on Water Bloom Prediction Based on Least Squares Support Vector Machine

An intelligent prediction model for water bloom of rivers and lakes based on least squares support vector machine (LSSVM) is proposed, in which main influence factor of outbreak of water bloom is analyzed by rough set theory first, and this model is compared with artificial neural network prediction model. The comparison result indicates: in the aspect of medium-term water bloom prediction in rivers and lakes, the accuracy of prediction with least squares support machine is higher than that of artificial neural network. Least squares support machine, which has long prediction period and high degree of prediction accuracy, needs a small amount of sample and can predict the medium-term change discipline of chlorophyll well. The results of simulation and application show that: LSSVM improves the algorithm of support vector machine (SVM)¿ it has long-term prediction period, strong generalization ability and high prediction accuracy; and this model provides an efficient new way for medium-term water bloom prediction.

Reconstructing indoor environmental 3D model using laser range scanners and omnidirectional camera

This paper present a method to acquire a realistic, visually convincing 3D model for indoor environment using a mobile robot platform with two laser range scanners and one omnidirectional camera. First, the vertical mounted laser scanner is used to acquire geometrical 3D model of indoor environment, while the horizontal mounted laser scanner is used to solve the simultaneous localization and mapping problem. After joint calibration of vertical mounted laser scanners and omnidirectional camera, we can combine the scan points of laser scanner with the corresponding pixels in the panoramic images, and then finally can get a high-quality real-time 3D model by incorporating the range data and color information. Experimental results from a real environment are presented at last.

Research on wastewater treatment system based on ADRC

Considering the characteristics of the time-delay, inertia and time-varying of DO control in the wastewater treatment, and on the basis of analyzing the defects on general PID control and Fuzzy Smith-PID control, one kind of ADRC controller is proposed in the wastewater treatment control system, Through the simulation experiment and comparing with general PID and Fuzzy Smith-PID control, the result shows that ADRC control strategy has lesser adjust-time and overshoot than other two control strategies. In addition, this control strategy not only can adapt the system change commendably under different working conditions, but also has good robustness and stability existing outside interference, and then provides an effective control way of solving the problem in the time-delay and inertia system.

Study on moving-objects detection technique in video surveillance system

A new method of moving-objects detection based on fusion of background subtraction and temporal differencing is proposed in this paper. The method constructs the adaptive background model by Gaussian model for each pixel in the image sequences, combined with temporal difference to update the background selectively, and simultaneous used background subtraction method extract movement areas from the background model. Then integration the two foreground regions segmented for object recognition, together with utilization of the median filter and mathematical morphology operation to eliminate noise and the small area of non-human movement parts. Finally obtain the complete reliable moving-objects. Experimental results show that the approach identifies targets accuracy, and can satisfy the needs of real-time in visual surveillance system.

A Control Method of Dissolved Oxygen in Sewage Treatment Based on Fuzzy-Smith

According to the characteristics of long time-delays, large inertia and time-variation in the dissolved oxygen control method of sewage treatment, one Fuzzy-Smith control model on dissolved oxygen control is presented in this paper. This model has the merits of a high robustness in fuzzy controller and a compensation for time delay in Smith predictor. After simulation experiment, the result is shown that this control model has both good robustness and stability when it matches the model or not, which is provided an effective approach to solve time-variation system of long time-delays and large inertia.

Neural network water bloom short time forecast based on evidence theory

Analyzing the characters of water-bloom eruption, one effective model on weightings attribute of forecasting water-bloom based on D-S evidence theory has been proposed. After pre-treating forecast index data, sets up water -bloom short-time forecast model based on neural network, which improves forecast precision of water-bloom, through simulation and testing, the result shows its affectivity and superiority.

Application of gray correlation analysis in eutrophication evaluative of water bloom

The monitoring data for water quality of Beijing Beihai between May and October in 2007 have been analyzed with gray correlation. In order to overcome the defects of the traditional method, this way consider the interzone form of the water quality evaluation standard, use centralization dimensionless functions, and combine the integrated nutritional status index with const weight. The thread system can reflect multi factors and is easy to use. The results showed that the improved gray correlation analysis is an active tool in environmental assessment and administration.

Prediction technique for water-bloom in lakes based on elman network

The outbreak of water-bloom is the result of coactions of water bodys physical, chemical, biologic and other progresses. It is very difficult to establish uniform mathematical model to efficiently evaluate and predict the water-bloom because of the water bodys biodiversity and nonlinearity. Based on the foundation of research in the mechanism of water-bloom and the main component analysis in rivers and lakes, modeling method, fundamental theory, principle and technique established for water-bloom prewarning system which is based on artificial neural network is specially researched. Elman network has the characters of good dynamic characteristics, fast approaching speed, high degree of accuracy and so on. In this paper, combined with the basic features of outbreak of water-bloom, Elman network is studied from the angles of theory and experiment and a water-bloom prewarning system in short term based on Elman network is established. Considering the defect of classical BP algorithm, improved BP algorithm is used for the training and study of network. Example analysis shows that Elman network model established in this paper is reliable and practical. Moreover, water-bloom prewarning system in short term in long river water system is also established by MATLAB in this paper which proposes a new method for intelligent research in water-bloom prewarning.

Optimization of water quality monitoring section based on comprehensive hierarchical clustering

In order to optimize the section layout of water quality monitoring, this paper proposes a new method based on comprehensive hierarchical clustering (CHC). Firstly, the method calculated the affinity-disaffinity relationship among the monitoring variables through 5 distance algorithms. Afterwards, the data set could be clustered automatically through 4 connection algorithms. Then taking the correlation coefficient as evaluation criteria, optimal hierarchical clustering algorithm was selected. Finally, with the corresponding optimal clustering tree matrix, the monitoring sections can be set optimally. In addition, the paper used students t test to verify the result of optimization. The experimental results show that this method can reflect the water quality of whole area more efficiently, thus has good prospect.

Research on Water Quality Monitoring Section Optimization Based on Multi-agent Model

In order to conduct reasonable and effective water quality (WQ) section monitoring of lakes, water reservoirs, and rivers this paper presents a multiagent model to optimize the WQ monitoring sections. First, the paper establishes a normalized matrix based on the original WQ monitoring data and from the data extracted principal/nonprincipal components as effective WQ features. Then, multi-WQ-agents model is built including comprehensive evaluation scores (CES) of principal/nonprincipal components in each WQ-agent and interaction rules of WQ-agents on the basis of the multiagent theory, in which the adjacent WQ-agents can be merged or split according to the similarity of CES. Therefore, the research realizes merge in the coarse segmentation of all WQ-agents, called as Agent # . On the condition that the number of Agent # s is smaller than the predefined threshold, the Agent # s would be spilt further in the fine segmentation, called as Agent*s. Finally, the central points of Agent* are selected as the measuring samples of WQ monitoring sections. The result shows that multiagent model can improve the monitoring quality, cut cost, and provides a creative measure of WQ monitoring section optimization.