Intelligent Microscope Imaging And Image Recognition Based On Jetson Nano

With the development of micro research in electronics,biology and medicine,the miniaturization of operating objects is driving the development of the microscope system towards automation and intelligence.Especially in the field of biomedical engineering,cell fusion,cell disease detection and pathological analysis need to be done with the help of a microscope.However,the traditional microscope requires manual focus to make pathological analysis of the observed target.This artificial way of working efficiency is very low,and due to human fatigue,subjective judgment and other factors may appear wrong,resulting in misdiagnosis or missed diagnosis.Jetson Nano is a powerful and low-cost artificial intelligence computing platform developed by Nvidia.Therefore,with the aim of developing low-cost artificial intelligence microscope,this paper studied artificial intelligence microscopic imaging and image recognition technology based on Jetson Nano platform.The intelligent microscope system consists of two parts: the automatic focusing system and the target recognition system.Firstly,the key structure of the intelligent microscope is studied,and the overall structure of the micro-operating system based on Jetson Nano is designed.Aiming at the problem of small imaging size of the micro system,based on the principle of super resolution microscopic imaging,the whole structure of the micro positioning platform with nanoscale positioning accuracy was designed by using the piezoelectric ceramic driving method and the flexible hinge guiding mechanism,and the statics analysis and finite element analysis were carried out.Secondly,part for microscopic imaging,image clarity evaluation algorithm is the key technology to realize automatic focusing based on digital image,and classical algorithm are susceptible to noise,illumination is not equal factors,low precision,poor stability,aiming at this problem,this article through the study of all kinds of classic image sharpness evaluation algorithm,convergence variance function and Brenner function,this paper proposes a new image sharpness evaluation algorithm and this algorithm is verified by MATLAB simulation in terms of sensitivity,stability,speed of significant advantage;For classic mountain climbing search algorithm is prone to local minima and slow problem,based on the variable step size mountain climbing algorithm proposed three steps,heuristic search method,away from the focus area adopts step search,near the focal region with small steps searching,near the extreme value point in the search,and through the forward to collect more two frames to compare with the size of the previous frame determine whether for local extremum,this article search strategy to avoid falling into local extremum problem,improve the search speed and precision.Finally,the image recognition part is mainly to focus the image content recognition classification,target recognition problem,high-precision,real-time and intelligent based on Jetson Nano studiedneural network platform,use the YOLOv4-tiny convolution homemade cells of neural network training data set,and use the Dropout technology,design the target identification model of this article DYOLOv4-tiny effectively avoid the fitting problem,improved the precision of recognition,through the experimental analysis shows that the system can accurately the content of the cell image recognition.Combined with artificial intelligence technology,this paper mainly studies and optimizes the fast focusing algorithm and target recognition method of the intelligent microscopic system.The research results have important reference significance for the development of the intelligent microscopic imaging analysis system.