Cheng Bo Yu’s research while affiliated with Chongqing University of Technology and other places

To process corneal endothelial cell image quickly and efficiently, observe the corneal endothelial cell form and relevant calculation and analysis. This system adopts mixed programming of LabVIEW and MATLAB, to have image processing and analysis of the corneal endothelial cell. Use LabVIEW to design operation and display interface, and send the images collected to MATLAB for complex arithmetical operation, then put out data in LabVIEW. The experiment indicates the system has the advantage of intuitive interface, convenient operation, and quick arithmetic, has the practical reference value.

Based on the current development of corneal endothelial cells in the domestic, this paper developed corneal endothelial cell analyzer software system which taken Microsoft Visual Studio as development platform in Windows7 system and combined with MATLAB image processing technology and the Access database information management technology. The software relies on a database established patient archiving and query system, by VC + + statically call the dynamic link library that MATLAB Compiler generated to complete image processing, by partial equilibrium approach for image enhancement, by dynamic difference method for determination threshold when binaries. The software provides good interactive interface and graphics display surface, but also sets aside network port as follow-platform extension.

In this paper, we study an an algorithm for accurately extracting the outline of Corneal Endothelial Cell images according to the structure and feature of the corneal endothelial cells. Firstly, we reduce the influence of uneven exposure to segmentation effect using histogram equalization. Secondly, reduce the image noise by Gaussian filtering. At last, research the classical algorithm of image segmentation through the contrast experiment and then select a simple and effective local NiBlack dynamic threshold algorithm. The experimental result shows that this processing method is not only simple, but also can segment the corneal endothelial cell images clearly, and provides a good foundation of data for accurate identification of the following images.

the optimal solution of weights can be solved through the mathematical programming model. The precision of weapon estimating can be improved via substitution of the solution for the subjective or objective weights which are regularly used in system estimate methods.

This paper puts forward a finger vein classification algorithm which combines Principal Component Analysis (PCA) with Radial Basis Function (RBF) neural network algorithm, named the PCA-RBF algorithm. Use the training sample to reduce PCA dimensions, and abstract the main component of the image. Because of the advantages of RBF neural network classifying, put finger vein images into different classes, and then use the shortest distance to recognize. Through the experiment result comparing with Back Propagation (BP) neural network, PCA-RBF neural network is better in finger vein recognition. The result shows that PCA-RBF has faster training speed, simpler algorithm and higher recognition rate.

Because of the DV-Hop algorithm has a big error in the estimation of the average hop distance, this paper proposed a weighted hop distance correction localization algorithm. The improved algorithm is carried out by introducing the average hop distance error correction value of the weighting processing, thereby reducing the hop distance error, and avoid the accumulation of errors in the subsequent computation process. The simulation results show that the improved DV-Hop algorithm reduces localization error effectively and has good stability without additional devices; therefore, it is a practical localization solution for WSN.

Design and implement an energy-efficient smart camera mote architecture to be used as surveillance device for assisted living. Add the Passive Infrared Sensor (PIR) to WVSN, PIR detect the human or animal’s moving, then it triggers the camera to wake up. The image captured will be grayscale processing by the central processing unit. Camera sensor nodes transmit a grayscale image over wireless channel to master control station. It offers reduced complexity, response time, and power consumption over conventional solutions. By experimental results from the test illustrate that performance of the designed wireless image sensor is better than the exiting ones in the market in terms of received signal strength intensity (RSSI) and packet rate ratio (PRR) with respect to the distance. This scheme is less complicated than other wireless video sensor surveillance techniques, allowing resource-constrained video sensors to operate more reliably and longer.

Design and implement an energy-efficient smart camera mote architecture to be used as surveillance device for assisted living. Add the Passive Infrared Sensor (PIR) to WVSN, PIR detect the human or animal's moving, then it triggers the camera to wake up. The image captured will be grayscale processing by the central processing unit. Camera sensor nodes transmit a grayscale image over wireless channel to master control station. It offers reduced complexity, response time, and power consumption over conventional solutions. By experimental results from the test illustrate that performance of the designed wireless image sensor is better than the exiting ones in the market in terms of received signal strength intensity (RSSI) and packet rate ratio (PRR) with respect to the distance. This scheme is less complicated than other wireless video sensor surveillance techniques, allowing resource-constrained video sensors to operate more reliably and longer.

It is difficult to have good performance to control large delay time system. A neural network identification method for nonlinear system’s delay time was discussed. Using the abrupt mutation resulted from the training error sum square of the real output and the expected output of the network, this method changed the input sample period of the neural network so that it could discriminate the delay time of the nonlinear model. Combining the discrimination of neural network system with long time delay and the control method based on model prediction, searching PID controller parameters based on ant colony optimization algorithm, it was applied to control boiler combustion system. The simulation results show that this scheme has much better advantage of celerity and robustness.

The transmitted intensity and the modulation depth of plane polarized light are implemented to investigate influences of the low frequency external magnetic field on the ferrofluid film. The frequency of AC magnetic field is modulated from 5Hz to 400Hz, and the thickness of ferrofluid films are changed as 12 μm, 25 μm, 140 μm and 520 μm, and the mass concentration of magnetic nanoparticles is 5 %, 10 %, 20 % and 33.6 %, respectively. The experimental results show that (1) the mean value of the optical transmittance of ferrofluid film is decided by the thickness of ferrofluid film and the mass concentration of nanoparticles, but the frequency of magnetic field has no direct relation with it; (2) under a given magnetic field intensity, the optical modulation depth of ferrofluid film is determined by the film thickness, the nanoparticle concentration and the magnetic field frequency. It is further found that the single relaxation characteristics of magnetic nanoparticle, whose size is smaller than the critical value of the single magnetic domain, is the important factor of the frequency properties of nanoparticle in ferrofluid. These results may be beneficial to applications of ferrofluid in magneto-optic sensors.