Please use this identifier to cite or link to this item:
Title: 運用波前修正於數位全像造影及其深度學習致動粒子偵測之研究
Studies on Wavefront Correction for Digital Holographic Imaging and Its Application in Deep Learning-enabled Particle Detection
Authors: 鄭超仁
Cheng, Chau-Jern
Gao, Yang-Jie
Keywords: 全像術
Digital Holography
Wavefront Correction
Zernike Polynomials
Deep Learning
Convolution Neural Network
Particle Detection
Issue Date: 2020
Abstract: 本論文主要探討利用數位全像式的資料及波前修正技術於深度學習以影像辨識上的優勢,以達到三維粒子偵測之目的。在數位全像造影中,本文探討波前像差對於樣品資訊的影響及修正方法,以得到正確的物體資訊,同時運用數位全像資料擴增方法,來提升數據集的多樣性。而運用上述方法即可透過數位全像術取得粒子的波前繞射資訊,再運用深度學習於物件偵測的技術,藉由調整模型架構及參數,來使樣品偵測能力及辨識能力達到最大準確度,來進行三維空間位置定位及尺寸分類,以利未來透過數位全像顯微造影系統擷取其他樣品的光場資訊進行定位,增加未來應用的潛力。
This thesis mainly discusses the advantages of using digital holographic data and wavefront correction technology in deep learning and image recognition for 3D particle detection. This article discusses the influence of wavefront aberration on sample information and correction methods to obtain correct object information. To diversify the data set, digital holographic data amplification methods are used. Using digital holography, the wavefront diffraction information of the particles can be obtained. Using this diversified data set, deep learning technology in object detection is applied. The model structure and parameters are adjusted to maximize the sample detection and identification capabilities. Accuracy is used for three-dimensional spatial location positioning and size classification. For future purpose, the light field information ofother samples can also be captured through the digital holographic microscopy system, increasing the potential for future applications.
Other Identifiers: G060777001H
Appears in Collections:學位論文

Files in This Item:
There are no files associated with this item.

Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.