Fast Focusing Method Of Microscope Based On Focal Plane Estimation

With the maturity and development of image technology,the convenience brought by automatic microscope technology has been recognized by various industries.In the task of observing the micro world,the microscope autofocus technology saves workers from repetitive and tedious tasks such as focusing and moving the field of view by using image algorithms and search strategies.However,there are still efficiency problems during the autofocusing and scanning of the microscope.Firstly,the resolution of images taken by microscope cameras has exceeded one million,and 2K images have gradually become mainstream,and the image itself has a large amount of calculation,resulting in a low efficiency of the evaluation algorithm for sharpness and aff ecting the focusing efficiency.Secondly,microscope imaging is susceptible to the influence of the light environment.The slight change in light causes the existing sharpn ess evaluation function curve to be non-monotonic,which affects the accuracy of the focusing algorithm.The existing sharpness evaluation function cannot simultaneously meet high computing efficiency and adaptively.Light intensity.Finally,the number of tasks that require image digitization is constantly increasing.There is still much room for improvement in the existing focus search strategies and scanning efficiency.Inefficient focus search strategies and scanning strategies will generate excessive searches,affecting the focus and scanning performance.In order to improve the efficiency of microscope autofocus,this paper improves the focusing efficiency from three aspects: sharpness evaluation function,focus search and multi-field scanning.The details are as follows:1.A fast focusing method of microscope with adaptive light intensity is proposed.In this method,the adaptive content position windowing method is proposed in terms of improving the calculation efficiency,which can add windows at the locations where the focused content is distributed densely,reducing the calculation amount of the definition evaluation function.In terms of sharpness evaluation function,a sharpness evaluation function of adaptive light intensity is proposed,so that a single-peak sharpness curve can also be generated under a changing light intensity environment.In terms of search strategy,a step-variable hill-climbing method based on the sharpness change rate was proposed to optimize the number of unnecessary focus times near the focus and improve focus search efficiency.2.A fast scanning method for focal plane estimation is proposed.This method proposes a method for identifying focused content based on gray histograms,which can determine whether there is focused content in the current field of view before focused search,and can save unnecessary focused search in fields of view without focused content.In terms of scanning,a scanning method based on focal plane estimation is proposed.The focal positions of oth er fields of view are estimated based on a known focused field of view,and different scanning strategies are used according to the needs of different focusing tasks to meet the balance of sharpness and scanning efficiency.3.Propose a regression-based microscope focus prediction method.This method learns a regression model based on the characteristics of different degr ees of focus in the same field of view.Using this model,the distance from the focus can be estimated.When focusing at a long distance,it can save the number of focus search times in large steps and improve the focusing efficiency.Experiments show that the windowing method proposed in this paper can adaptively focus the content position,and accurate windowing reduces the amount of calculation.The proposed sharpness evaluation function can adapt the illumination intensity while maintaining the calculat ion efficiency.The proposed scanning method based on focal plane estimation can efficiently scan the whole slide.The regression-based focus prediction method proposed in this paper can quickly find the near focus at the far focus and estimate the focus p osition at the near focus.