Research On Vision-base Methods For Quality Measurement And Selection Of Microspheres

Microspheres with diameters ranging from tens to hundres of micrometers have extensive applications in laser fusion, photonics, pharmaceutical assays and so on. Geometric parameters and pressure resistance are crucial attributes to determine the qualities of microspheres and need to be strictly controlled. At present, the quality measure of microspheres is carried out manuanlly, the precision and proficiency of which are limited. Morever, traditional vacuum and mechanical grippers would do harm to microspheres' surface during the manipulation of qualified microspheres. In this study, we propose vision-based quality measurement and compliant manipulation.For the quality measurement of microspheres, geometric parameters including radii, roundness and deficiency as well as pressure resistance are determined. Canny detector is used to obtain edge images, which would undertake gradient Hough transform for circle dectection. Ring regions of interest are employed for image segmentation, then extracted contours of microspheres are fitted to ellipses via least square fitting. For the pressure resistance measurement of microspheres, we use image matching based on Hu invariant moments to match the images under pressure with the images of intact microspheres before pressure loading, by which the failure fo microspheres could be detected when the matching values exceed the threshold.Compliant selection of microspheres is implemented under the attractive force exerted on microspheres by liquid bridges, which are formed as the condensed droplets on the hydrophobic probe contact the microspheres. We propose a vision-based method to identify the liquid bridge and extract its contour. By solving Young-Laplace Equation with fitting contour equtions, one can get real-time force feedback from the liquid bridge. Besides, by regulating condensation conditions, the force from the liquid bridge could be controlled for better maneuverability. We employ an image-based visual servoing scheme in which pyramid Lucas-Kanade optical flow is used to track the probe and provide visual feedback. Normalized variance is choosen to be the measure of focus and facilitate autofcocusing of the probe. Lastly, inertial release strategy is adopted based on vibration of piezoelectric material, overcoming the adhesion force between the probe and the microsphere.To verify the performance of proposed methods, customized experimental platforms were set up. In experiments for measurement of geometric parameters, 100% successful detection rate was achieved and the measurement speed reached 100 microspheres per minute with average radii error as 4.7 ?m. Micropheres could also be classified based on the measurement of roundness and deficiency. As for the measurement for pressure resistance, all failures of microspheres under pressure were correctly discrimated and according pressures were automatedly recorded with little delay. Compliant selection tests were conducted and experimental results indicated that microspheres with 200-800 ?m diameters were reliably grasped by liquid bridges and released to target positions with inertial release strategy. The error for force measurement of liquid bridge is 1.53 ?N, and positoning error of visual servoing is about 8 pixels. Experimental results validated the robustness and accuracy of vision-based quality measurement and selection of microspheres. Compliant selection for qualified microspheres could guarantee the capabilities of fusion targets.