2013-08-012024-05-18https://scholars.lib.ntu.edu.tw/handle/123456789/702594摘要：生物藉由繁衍時遺傳基因的變異改變其外型。花朵的形狀複雜，且其變異多端。因此花朵外型的變異一直以來都是生命科學家與生物工程師關注的課題。傳統的方法大多以二維的方式描述花朵的幾何形態與分析其變異。然而，花朵是三維物體，傳統的方法不足以完整地捕捉花朵外型的幾何資訊。本研究提出以三維的方式建立花朵的幾何模型，並發展幾何形態測量工具以分析花朵形狀變異。三維花朵幾何模型的建構會利用立體機器視覺和微米級電腦斷層掃描造影技術。立體機器視覺的方法將會被應用於建立結構單純花朵的模型。立體機器視覺使用兩個以上的攝影機，在不同的視角捕捉多張花朵的二維影像。利用花朵與攝影機之間的幾何關係，即可推算其各部位在三維空間中的座標，建立其三維模型。微米級電腦斷層掃描造影的方法將會被應用於建立結構複雜花朵的模型。電腦斷層掃描造影在不同的角度，對花朵發射X射線，測得花朵樣本的多個X射線二維影像，再利用背投影的方式重組花朵的三維模型。獲得三維花朵模型後，本研究便會開發幾何形態測量演算法，使用電腦自動補捉花朵的幾何資訊，並利用泛用型普式分析與主成分分析，量化花朵的形狀變化與特徵。本研究的結果可以做為花卉產業產品形狀分類的工具，並可被應用在未來的研究，尋找花朵基因變異和外型表現之間的關聯。<br> Abstract: Creatures change their appearances and shapes from generation to generation through mutation. Floral shape is sophisticated, and the variation in shape is extensive. The functional consequences in the shape of flowers have intrigued scientists and bio-engineers for centuries. Traditionally, two-dimensional (2D) methods are applied for floral shape modeling and variation analysis. However, flowers are essentially three-dimensional (3D). The traditional 2D methods are not able to adequately capture all the geometric information of flowers. This research proposes to model floral shape in 3D and to develop geometric morphometric tools for studying floral shape variation. The 3D floral shape modeling will be carried out through stereo machine vision (SMV) and micro-computed tomography (micro-CT). The SMV approach will be used to model structurally simple flowers. The SMV uses two or more cameras to capture 2D images of a flower specimen at different viewpoints. The geometric relationship between the specimen and the cameras are used to construct a 3D flower model from the 2D images. The micro-CT approach will be used to model structurally complex flowers. The micro-CT emits X-ray that penetrates a flower specimen at different angles and records the X-ray signal using a detector. The X-ray signal of the specimen is reorganized into a 3D model via back projection. With given the models, algorithms will be developed to characterize the geometric information from the 3D flower models automatically. Then generalized Procrustes analysis and principal component analysis will be performed with the geometric information to quantify features in floral shape variation. The results from this work can be a powerful tool for the domestic floral industry in classifying shapes of flower products. The results will also be applied in future studies, aiming to find the association between genotypic and phenotypic variation of flowers.三維模型幾何形態測量學花朵幾何形狀變異立體機器視覺微米級電腦斷層掃描造影主成分分析泛用型普式分析表型基因型Three-dimensional modelingGeometric morphometricsFloral shape variationStereo machine visionMicro-computed tomographyPrincipal component analysisGeneralized Procrustes analysisPhenotypeGenotype