Research On Parameter Detection And Transfer Control Method Of Microspheres Based On Microvision

In the field of controlled nuclear fusion,the diameter of the hollow microspheres used to load the fusion fuel is usually between two hundred and two thousand micrometers.The diameter and surface quality of microspheres have a crucial effect on the fusion experiment and need to be strictly controlled.At present,the selection of large quantities of microspheres is mainly based on manual detection,which has low efficiency and cannot meet the accuracy requirements.In this paper,a parameter detection and transfer control method of microspheres based on micro vision is proposed,which can realize the automatic parameter detection and lossless transfer of microspheres and improve the precision and efficiency of preliminary screening of microsphere.In order to detect the diameter of microsphere,the micro vision technique is used to identify the boundaries of microsphere.After gray level transformation and Gauss filtering,the contrast limited adaptive histogram equalization is used to enhance image,so as to improve the contrast between the microspheres and the background image.According to the diameter and position of microsphere in the rough detection results of gradient Hough transform,the seed region is set up for region filling and contour extraction to eliminate the inner boundary of the microsphere.The edge of the microsphere is extracted by an adaptive region of interest.The least square ellipse fitting of the boundary points is carried out to obtain the precise diameter and position information of the microspheres,and the automatic boundary recognition of the microspheres is realized.According to the defects of the surface of microspheres,the linear classification method based on feature extraction and the nonlinear classification method based on neural network were used to divide into three types: smooth,rough and malformed.For the linear method based on feature extraction,the grey distribution function and cumulative distribution function are calculated on the premise of extracting the information of the image inside the microsphere.The piecewise linear fitting of the cumulative distribution function after normalization is carried out.Two characteristic parameters,uniformity and transparency,are extracted from the fitting function to describe the surface quality of microspheres quantitatively and a linear classifier is designed according to these two parameters to realize the defect segmentation of microspheres.As for the nonlinear classification method based on neural network,a three-layer back propagation neural network model is built,whose input signals are the gray levels of the internal image of the microspheres and image background.The model is optimized by training set and verification set and the final optimized model is used as a nonlinear classifier to realize the defect of microspheres.For the lossless transfer of microspheres,the vacuum absorption clamping method is adopted.The kinematics equation of micro-gripper picking operation is established according to the position-based visual servo control method and the kinematics equation of micro-gripper release operation is established according to the structure of the plate.The system parameters in these two equations are calculated by the parameter calibration program.Two operating modes,single selection and cycle selection,are designed to realize automatic detection and transfer of microspheres on the microsphere operating platform.In order to verify the performance of parameter detection and transfer control method of microspheres,relevant experiments were designed.In the experiment of boundary recognition,the average measurement error of diameter is 7.81?m,the average repeatability error is 2.81?m,and the number of recognition per minute is about 53.In the experiment of defect detection and classification,the classification accuracies of the both classification methods were above 85%,but the accuracy of linear classification method based on feature extraction is higher and more suitable for rough selection work of microspheres.In the lossless transfer experiment,the average position error of the left micro-gripper in the visual servo system is 12.72?m,and the average position error of the right micro-gripper is 11.61?m.There was no damage during the transfer of microspheres and the successful rate of transfer was 83.87%.The transfer speed of single micro-gripper was 40 seconds each,and the coordinated transfer speed was 30 seconds each.The experimental results verify the accuracy and efficiency of the parameter detection and transfer control method of microspheres based on micro vision,and show the advantages of the method of automatic detection and selection of microsphere operation relative to manual selection.