玉米果穗DUS性状测试的图像处理应用研究

 【目的】评价图像处理法采集和数量化玉米果穗特异性、一致性和稳定性测试（DUS）性状的技术适用性。【方法】以4个品种各50个果穗及8个品种93～107个穗轴为材料,通过图像处理采集玉米DUS测试指南规定的7个性状（穗长、穗粗、穗形、粒顶色、穗轴色、穗行数和籽粒排列形式）,应用多性状整体控制单一比较法分析品种特异性。【结果】穗长、穗粗和穗行数的图像处理误差分别为6.2%、1.6%和0.66%。果穗的穗缘角（穗形）在0～2.22°之间变化,各品种穗行角（籽粒排列形式）均值在89.4°～90.7°之间。穗形等4个质量或假质量性状成功转换为数量性状,信息量随之增加。籽粒顶端颜色在果穗间和果穗侧面间的差异都极其微小,其它性状的果穗侧面间差异与果穗间差异相当或者更小。图像处理容易获得同源样品的更多形态性状,可能导致品种伪差异的风险升高。【结论】图像处理具有客观、高效、低成本地采集和数量化玉米果穗DUS性状和其它更多性状的能力,结合多性状整体控制单一比较法等适当的统计分析技术,将在中国的新品种DUS测试中发挥越来越重要的作用。

Study on Application of Image Process in Ear Traits for DUS Testing in Maize

Objective] The objective of this study is to assess the suitability of image process techniques for measuring and quantifying ear traits in maize DUS tests. Method] The 7 ear traits, namely ear length and width, kernel top and ear axis color, ear shape, ear row number, kernel arrangement, as ruled in The National Guidelines for Maize, were measured by image processes from 50 ears each of four cultivars or 93 up to 107 ear axes each of eight cultivars. Measurements obtained were subjected to different statistical procedures in order to determine adequate data analysis requirements, in the context that the variety evaluation of distinctness. Result] Relative measurement errors were 6.2%, 1.6% and 0.66% respectively for ear length, ear width and ear row number by image process. Individual ear edge angles, as a trait depicting ear shape, ranged from 0 to 2.22 degrees, and mean kernel row angles, as a trait depicting kernel arrangement, varied from 89.4-90.7 degrees among cultivars. Colors and shapes of ears, which, as usual, are quality traits or pseudo-quality traits, were quantified as quantitative traits so that information gain increased. Variations in kernel top color among both ears and sides within ears were tiny, and variations in other traits among sides within ears were smaller than or comparative to among ears. The increased number of traits based on same bulk samples may lead to higher risk of false between-varieties distinctness. Nevertheless, the risk of this kind can substantially be reduced by the mean separation for individual traits with multiple traits adjustment. Conclusion] Image process is a useful tool for gathering and quantifying maize ear DUS and other more waits with advantages of objectivity, efficiency and low cost, when integrated with adequate statistical analysis tools such as the mean separation for individual traits with multiple waits adjustment, and will play more and more important roles in the new maize variety DUS testing in the whole country.