| With the rapid development of life science,more and more researches on cell clone,test-tube baby,drug research and development,cell segmentation based on cell micromanipulation have been carried out,which have solved many medical problems.Traditional manual cell micromanipulation requires skilled professionals.Even so,it still has low efficiency,poor repeatability and low cell viability.In order to meet the biological needs of large quantities,the automation of micromanipulation is an inevitable trend of development.At the same time,machine vision is applied to the micromanipulation system.It can replace human eyes to obtain cell location,actuator end tool location and other information,which is an indispensable part of the automated micromanipulation system.At present,the image processing speed in the automatic micromanipulation system is slow,which cannot provide real-time visual feedback information for the servo control system,so the control accuracy is low.In this paper,a visual cell micromanipulation system is built to solve the above problems.I design a robust and real-time image processing system and visual control system.In the paper i analyze the general steps of cell injection to obtain the design requirements of cell automatic injection,and build a micromanipulation system suitable for cell injection.The system includes hardware system and software system.Hardware system is divided into vision module,control module and actuator module according to different functions.Each module is introduced about function introduction and model selection design in detail.The software system introduces the platform environment of software construction,At the same time,each hardware module develops the appropriate software functions.In the image processing system,the automatic focusing algorithm based on image is researched,we obtain the relative depth between cell and probe,we also obtain clear images,the algorithm improves the automation of the system.In this paper,five commonly used definition evaluation functions are analyzed,and the Tenengrad sharpness function is improved.The more sensitive K-Tenengrad sharpness function is proposed.In the aspect of probe recognition and tracking,a template matching algorithm based on contour features is used,we take the gradient of image as the contour feature.Adding the punishment mechanism and designing the appropriate scoring mechanism improve the stability of the algorithm.Image gradient direction has binary operation,image pyramid combined with breadth-first search strategy and greedy early termination search method can speed up the algorithm.A particle filter algorithm based on shape feature is proposed to solve the problem that the classical particle filter based on color space.The classical particle filter is sensitive to illumination change and the color of tracking area is similar to the background.Using shape feature template matching instead of color histogram matching to calculate similarity,the algorithm is more stable when the illumination and color are similar.The micro operating system only obtains external information through vision,so it adopts visual servo control based on image and controls the mechanical arm through PID controller.5 groups of zebrafish embryo cells were injected with automatic microinjection.The feasibility of the micro operating system was verified by cell injection experiment. |
