Research On Visual Servo And Motion Control System Of Biological Automated Microinjection Process

Biological microinjection plays a very important role in the fields of toxicology research,gene editing,and drug testing.Traditional manual microinjection has problems such as poor accuracy,low success rate,and easy contamination of biological samples.Based on this background,the research on bio-automated microinjection systems has been valued by many countries in the world.China has also emphasized the development of comprehensive cross-industry in the "National Key R & D Plan",and has clearly focused on the research of Micro-operation technology.Compared with human genes,zebrafish have higher genetic similarity,and have become the main model for studying human diseases and tumors.Based on the existing research,this paper combines the image algorithm and deep learning model to identify the zebrafish's three-axis pose.At the same time,study the posture adjustment strategy and cooperate with the designed rotating mechanism to adjust the zebrafish to the best injection posture.In addition,aiming at the outliers caused by the occlusion of the zebrafish body surface pigments,an injection control algorithm based on adaptive robust Kalman filtering is studied to achieve high precision and high survival rate automated microinjection.This article mainly does the following work:Firstly,an automated microinjection system framework is designed for the zebrafish larval heart.Unlike traditional microinjection of spherical or spheroidal samples such as cells or embryos,the fixation and posture adjustment of zebrafish larvae are difficult.This article designs a set of three-axis rotation equipment for zebrafish and other shuttle-shaped samples,including a horizontal rotation radio and a body-axis rotation actuator,which can adjust the zebrafish's three-axis posture.Combined with micro-nano equipment to realize the automatic microinjection of zebrafish larval heart.Secondly,study image algorithms and deep learning model of zebrafish microinjection system.Compare the binary area of the current image and the template to determine if the zebrafish is in the field of view.When the zebrafish is not fully exposed,the image moment is used to obtain its coordinate information,and it is coordinated with the XY radio station to move it to the fully exposed area.In the case of full zebrafish exposure,image algorithms and deep learning models were used to identify the zebrafish's level and body axis posture,and the designed posture adjustment device was used to rotate the zebrafish to the optimal injection posture.Based on prior experience,a P-Tile segmentation algorithm was used to obtain a binary image of zebrafish eyes separated from fish stings.Based on this,a region of heart of interest was constructed,and a cumulative detection method was used to highlight the beating heart.The pyramid template matching algorithm was used to obtain the horizontal coordinates of the end of the actuator,and the contact method was used to obtain the relative height of the Z axis of the end of the actuator.Thirdly,the motion control algorithm of zebrafish microinjection system is studied.According to the horizontal pose information of the zebrafish,it is divided into the horizontal position and the tilt angle region.Different areas correspond to different inputs of the PID controller,which respectively control the total exposure and horizontal pose adjustment.Research on the position-based visual servo system to adjust the zebrafish's horizontal pose without the need to ensure that the zebrafish is always in the microscopic field of view.At the same time,the centrifugal compensation is used to suppress the centrifugal drift movement during the rotation process to ensure that the zebrafish finally falls in the field of view after servo adjustment based on coordinate transformation.In response to the phenomenon that the zebrafish occludes the injection needle during the injection process and produces abnormal values of the tip positioning,according to the state equation and measurement equation at the end of the injection needle,an adaptive robust Kalman filter is used to suppress the abnormal value of visual feedback and improve the injection.Angle and Zebrafish Survival.Finally,an automated microinjection system experiment is constructed to carry out related experiments.Experimental analysis and verification are performed from image algorithms,deep learning models,pose adjustment strategies,and injection control algorithms based on adaptive robust Kalman filtering.Experimental results show that the image algorithm and deep learning model can meet the requirements of larval pose recognition and accurate heart positioning.The posture adjustment strategy and designed rotating mechanism can complete the posture adjustment stably and quickly.The injection control based on adaptive robust Kalman filtering can effectively suppress abnormal values of visual feedback in real time,and can realize high-precision,high-survival automated microinjection of zebrafish heart.