2016-08-012024-05-18https://scholars.lib.ntu.edu.tw/handle/123456789/710303摘要：水稻是人類的主食之一。許多現存水稻品種具有重要的性狀。透過量化水稻表型特徵，與關聯性分析尋找表型基因座，成為有效提高水稻生產的方式之一。傳統的人工表型量化方法耗時費力且主觀；相對地來說，機器視覺方法即時、客觀，且非破壞。因此，本研究提出使用機器視覺量化不同品種稻穀和稻苗的表型特徵。此外，近年來由於食安問題頻傳，食品安全認證的需求也日益擴大。稻穀為大量交易的食物之一。本研究所提出的機器視覺特徵定量方法，也可以使用在稻穀品種的辨識上。本研究所提出的方法將被應用在「水稻多樣性種原」所收集的品種上。多樣性種原收集413種同基因合子水稻品種。這些品種廣泛涵蓋水稻遺傳變異，因此可成為研究水稻基因型和表型之間關聯性的優良材料。本研究的主要目標是：(1)擷取稻穀表型特徵，(2)量化稻穀的跨品種形狀變異，(3)識別稻穀品種，(4)利用全基因組關聯研究方法尋找稻穀表型基因座(如顏色或形狀)，(5)建立機器視覺影像系統以量化稻苗表型特徵，(6)量化稻苗農藝性狀(如耐鹽、耐旱，或抗病)，(7)利用全基因組關聯研究方法尋找稻苗表型基因座。本研究所提出的方法可望提高對稻穀和稻苗表型性狀量化的精確度，並提昇全基因組關聯研究的有效性。<br> Abstract: Rice (Oryza sativa L.) is a leading staple food worldwide. There exists thousands of rice cultivars, some of which develop important agronomic traits. It is desired to phenotype these traits, so that the genes associated with the traits can be identified subsequently to improve rice production. It is known that conventional phenotyping methods through naked-eye examinations are typically labor-intensive, time-consuming, and subjective. In contrast, image-based approaches are prompt, objective, and nondestructive. Hence, this study proposes to quantitatively phenotype the traits of rice grains and seedlings for various cultivars using machine vision. In addition, the demand for food authentication is increasing worldwide in recent years. Rice grains are traded globally in great amount. The proposed image-based phenotyping approaches can also be employed to authenticate the cultivars of rice grains. As a practice, the proposed approaches will be applied to the cultivars collected by rice diversity panel (RDP). The RDP comprises 413 purified homozygous rice accessions and represents a broad range of genetic variation in rice, hence serving as excellent materials for studying the association between the genotypes and phenotypes of rice. The single nucleotide polymorphisms of the RDP cultivars are also publicly available for the genome-wide association study (GWAS). The main objectives of this research are to (1) phenotype rice seeds of the RDP cultivars, (2) identify principal rice shape variation across the cultivars, (3) develop classifiers to recognize cultivars for the seeds, (4) perform GWAS to identify the loci associated with phenotypic variations (e.g., color or shape) of the seeds across the cultivars, (5) establish an imaging system to phenotype rice seedlings of the RDP cultivars, (6) quantify agronomic traits (e.g., resistance to salinity, drought, or disease) of the seedlings, (7) conduct GWAS to identify the loci associated with the phenotypic variations of the seedlings across the cultivars. The proposed approaches are expected to improve the efficiency and effectiveness in phenotyping rice seeds and seedlings, and to enhance the validity for the GWAS on rice traits.機器視覺影像處理機器學習幾何形態測量學形狀變化稻穀品種分類水稻生理數量性狀位點全基因組關聯研究machine visionimage processingmachine learninggeometric morphometrics (GM)shape variationrice cultivar classificationrice physiologygenome-wide association studyquantitative trait loci (QTLs)