Font Size: a A A

Automatic Focusing And Global Precise Imaging Of Pathological Microscope Based On Convolutional Neural Network

Posted on:2020-07-26Degree:MasterType:Thesis
Country:ChinaCandidate:Y H GeFull Text:PDF
GTID:2392330620459868Subject:Mechanical engineering
Abstract/Summary:PDF Full Text Request
This paper studies the algorithms of pathological microscope autofocus and global precision imaging.The proposed algorithms cooperate with the high-speed camera,focus motor and micro-motion platform to promote the ability of traditional pathological microscope.Achieving global target detection,slice scan planning,automatic focusing,defocus image restoration,automatic stitching of global imaging and other automated processing in the process of full slice scanning diagnosis of microscopy.This paper aims to solve the problems of the limited field of vision,low efficiency of manual focusing,low accuracy,poor image acquisition quality and low fault tolerance of traditional pathological microscopy.Targeting for the core steps of pathological microscopy for automated pathological slice scanning,the knowledge of mechanical electronics,computer vision,software engineering,and other disciplines were utilized in algorithm research and system design.A monocular camera was used to achieve global detection,design target detection algorithm based on machine vision,accomplish the location of scanning slice and light source,and comprehensive planning of scanning.The microimage sharpness evaluation algorithm was proposed based on convolution neural network and defocus estimation was accomplished by the classification algorithm.The algorithm drives the focusing motor to make precise automatic focusing on slice scanning.The Z-axis auto-focusing platform is redesigned with XY micro-motion platform,and the hardware circuit is designed.A cross-modality image mapping algorithm was applied based on the generative adversarial network to defocus microscopic image restoration.Aiming at the characteristics of microscopic image acquisition and stitching,a large-scale fast stitching algorithm based on feature matching is designed to improve the accuracy and efficiency of massive image stitching.This paper design the PC human-computer interaction software based on the hardware platform and control algorithms and achieve the integration of whole slice scanning,auto-focusing and precise imaging based on the microscope.This paper focuses on the design of the core algorithms for the automatic process of scanning,focusing and global imaging of microscopes.Qualitative and quantitative experiments were carried out to verify the accuracy and efficiency of auto-focusing,the effectiveness of image stitching and the quality of defocus image restoration.The critical algorithms are deeply excavated and migrated.The proposed algorithms have great application value in the field of accurate diagnosis.
Keywords/Search Tags:Pathological microscope, Convolutional Neural Network, Autofocus, Full slice scanning, Defocus restoration, Image stitching
PDF Full Text Request
Related items