Microscope Auto-focusing Method Based On Environmental Perception

Automated microscopy has a wide range of applications in a variety of industries,particularly in the medical field,where doctors use it to view tissue sections of patients and diagnose them based on medical images.This traditional method of manual reading not only brings heavy workload to doctors,but also results in misdiagnosis as a result of visual fatigue of doctors.With the development of image processing and artificial intelligence,automatic reading technology began to emerge.This technology first acquires a large number of continuous photos through a computer-controlled microscope moving platform and then analyzes the abnormal cells and diseased tissues through image processing and machine learning algorithms.This technique effectively reduces the working intensity of doctors and improves the diagnostic accuracy of doctors.The workflow of the existing automatic interpretation technology is to first move to a designated location,then automatically focus,capture the photos,and then move to the next location until the number of captured photos meets the requirements.As the sample slide and microscope camera is not an absolute vertical relationship,the platform will appear blurred in the process of moving the image,so the purpose of focusing after the movement is to ensure that the microscopic field of vision is clear.In recent years,the image resolution goes up to 2048×2048 with the rapid development of the camera,and the computing efficiency of traditional focusing methods are extremely low in high-definition images,which affects the scanning speed seriously.However,not all of the microscope fields are blurred in the actual scanning process.Focusing on a clear view is not only meaningless,but also wastes time for the overall scan.On the other hand,with the flatness of the electric platform constantly increasing,the adjacent focus position is likely to be in a plane.The conventional scanning method focuses on the known plane to reduce the overall scanning efficiency.In order to improve the efficiency of automatic scanning,this paper improves its efficiency from three aspects: focusing,scanning and focus estimation.The work is summarized as follws:First,this paper proposed an efficient auto-focus method.The method proposed gray-scale non-zero statistical function and low-gray-value statistical function in the aspect of sharpness evaluation function,which improved the current step-by-step hill-climbing method in focus search.The method utilizes the gray characteristic of the image and adopts the combination of two sharpness evaluation functions.According to the distance of the focus,three steps are adopted.This method compared with existing methods,it has higher accuracy,less focus time,and greater focus range.Next,there is a fast-focusing method based on image sharpness recognition.The method first uses a machine learning algorithm to prioritize sharpness prior to each autofocus and the image was judged weather it is clear,then the automatic focusing was decided.This method can effectively skip the sharp picture of the auto-focus,improve the overall efficiency of the scan.Finally,the paper mentioned a fast scanning method based on multi-point focusing.The method selects a few positions to focus,and determines the focus of each position by the method of plane fitting.And then platform was moved to the current location that in the center of the view to complete the picture collection.This method calculates the focus of the adjacent position with a few focal points,and does not need to search the focus of the mobile platform,which greatly improves the efficiency of the scanning.Experiments show that the new auto-focusing methods proposed in this paper perform better than existing focusing functions in sharpness focusing function and focusing speed.The fast scanning method based on image sharpness recognition can effectively improve the scanning speed and maximize the clarity of images.The multi-point focus quick scanning method can collect images with higher resolution than the original method,and improve the overall scanning efficiency.