Cao Shengxian

A colour constancy algorithm based on neural network and application

Obtaining color constancy is the study focuses in the field of image vision which does not alter with light source illumination modification. In consideration of the influence of the light source fluctuations on the image color, this paper puts forward an image color constancy algorithm based on BP neural network. This algorithm was used for testing different light source and illumination intensity in the same goal color which can achieve the non-linear mapping from the color values under different illumination polychromatic constant current light source to the color values under polychromatic standard light source. The speed of calculation is fast and the error is small for this algorithm. And this algorithm is applied to the hardness online measurement based on the image color. The experiment proves the effectiveness of the color constancy algorithm which can reduce or eliminate interference of the light source, correct color, improve the measurement sensitivity and accuracy.

Study of Circulating Cooling Water Fouling Based on Electronic Technology and Multi-parameter Detection

Scale sample of the iron content is an important symbol of the extent of corrosion to determine the circulating cooling water installations. In order to satisfy the projected need fouling features, the experimental design for dynamic closed loop cooling water system simulation of scale-like generation based on electronic technology and multi-parameter detection, and take dynamic water samples which iron content in real time during the simulation, microbial number、pH、dissolved oxygen, corrosion rate etc. In the precise analysis, JY/T015-1996 plasma emission spectroscopy measured in scale sample of iron ions, copper ions, organic matter content. The results show that the water has high iron, high-scale sample of the iron detection value indicates that the extent of corrosion is also relatively high; In addition, the copper ion content in the scale sample of high-lead content of organic matter, both a negative correlation.

An Edge Detection Binarization Method for Microbial Adhesion Image

This paper makes the image edge detection aiming to microbial adhesion image, and establishes a new image binarization method based on the uneven characteristics of microbial adhesion. This method makes the captured image gray-scale and median filtering processing, and then gets the edge detection image and histogram respectively. For histogram, this algorithm first finds the gray value which corresponds to the maximum frequency and divides the histogram into two intervals, then gets the edge threshold in each interval and makes the image binarization.The edge detection image of binary image is compared with the filtering edge detection, when the two images coinciding edge points are the most, the binary image at this time is the final binary image. This method has a good result comparing with local threshold method?global threshold method and dynamic threshold method.

Paper Prediction Model of Biofouling on GA-BP and Dynamic Simulation

The prediction model of cooling water biofouling was established that selected water quality factors, according to the magnitude of the effect of pH, conductivity, dissolved oxygen, total bacterial count on the fouling resistance and time as the input variable, take the fouling resistance as the output variable. The model combines genetic algorithm (GA) based on global optimization with back propagation (BP) based on gradient descent in the paper make the linking weights of networks self-adaptive evolution in constantly iterative process. The results show that GA-BP could significantly increased model computational speed and accuracy, proved the effectiveness and reliability of this method provides a new way for guiding the cooling water treatment and improve model accuracy.

Forecasting Corrosion Rate of Coolingwater Based on Least Squares Support Vector Machine

In view of the corrosion of cooling water system, the dynamic simulation test was conducted with the cooling water dynamic simulation experiment device. In the test period the corrosion rate and the water quality factors were monitored. Based on the test data, an intelligent prediction model of cooling water corrosion rate based on least squares support vector machine (LS-SVM) is constructed, in which the water quality factors related with corrosion were selected as input variables and the corrosion rate was selected as output variable. The results show that the LS-SVM model is pithily, and it has better extensive capability than traditional methods. The new method is effective and reliable, and it can be viewed as a new approach to advance the development of cooling water treatment technology and improve the prediction accuracy of the corrosion rate.