期 刊 ：植物遗传资源学报 2015年 16卷 4期
Keywords:sweetpotato, pigment, root flesh color, dry matter content, RGB characteristic value,
摘 要 ：为明确甘薯色素、干率与肉色的关系，给图像RGB特征值评估甘薯色素含量提供依据，测定了甘薯240份品系的胡萝卜素含量及234份品系的花青素含量、干率和薯肉RGB颜色特征值。结果表明，甘薯胡萝卜素含量与RGB值的相关系数r＝-0.913，偏相关系数为-0.872，花青素含量与RGB值的相关系数r＝-0.838，偏相关系数为-0.836，说明甘薯色素含量与肉色是高度正相关的，是肉色的决定因子，但胡萝卜素与肉色关系更紧密。干率与胡萝卜素RGB值的相关系数r = 0.597，偏相关系数为0.273，干率与花青素RGB值的相关系数r = -0.273，偏相关系数为0.255，说明干率与对肉色有极明显的负面影响，但是在RGB值≥200的低胡萝卜素（约含量3 mg/100g.FW以下）群体中，干率与RGB值的偏相关系数为-0.0066（P = 0.9426），干率对低胡萝卜素含量品系的肉色没有影响。采用R值、G值、B值、干率的二次多项式回归模型，RGB值≥200的低胡萝卜素群体、RGB值<200的较高或高胡萝卜素群体和花青素群体的拟合方程的复相关系数R分别为0.975、0.982和0.937，说明拟合值与实测值关系十分密切。对胡萝卜素含量高于3 mg/100g.FW的品种，拟合值能达到约90%的精确度，对胡萝卜素含量低于3 mg/100g.FW的品种，也能达到约83%的精确度，对花青素品种能达到约86%的精确度。因此，图像RGB特征值法不仅能够很好地反映甘薯品种间的色素差异，还能较精确地估测甘薯色素含量，可为甘薯早代育种的大批量样品提供便捷、客观的色素含量评估方法。
Abstract:In order to clear the relationship of pigment and dry matter content with flesh color of sweetpotato, and provide estimation methods for pigment content based on the RGB characteristic values of images, the carotene content in 240 strains, anthocyanin content in 234 strains, the relative dry matter content and RGB characteristic value were determinated. The results showed that the correlation coefficient of carotene content with RGB value was -0.913, and their partial correlation coefficient was -0.872, while the correlation coefficient of anthocyanin content with RGB value was -0.838, and their partial correlation coefficient was -0.836, indicating that pigment content was a determining factor and high positive correlation with root flesh color, however there was a closer relationship between carotene content and root flesh color. The correlation coefficient of dry matter content with carotene RGB value was 0.597, and their partial correlation coefficient was 0.273, while correlation coefficient of dry matter content with anthocyanin RGB value was -0.273, and their partial correlation coefficient was 0.255, indicating that dry matter content had very obviously negative effects. But in low carotene (about 3 mg/100gFW or less) group with the RGB value> 200, the partial correlation coefficient of dry matter content with RGB value was -0.0066 (P = 0.9426), so dry matter content had no effect on root flesh color of low carotene strains. The pigment content estimation achieved good fitting results using the quadratic polynomial regression model with R value, G value, B value, and dry matter. Multiple correlation coefficients of fitting equations in low carotene group with RGB values < 200, high or moderately high carotene group with RGB values ≥200, and anthocyanin group, were respectively 0.982, 0.975, and 0.937, indicating that the fitting values had close relationships with measured values, and their fitting values could reach respectively about the accuracy of 90%, 83%, and 86%. Therefore, the RGB characteristic value might be used as not only a good indicator of pigment content difference on varieties, but also a relatively accurate estimation for their pigment content, it could provide a convenient and objective evaluation method of pigment content for large sums of samples in the early-generation selection of sweetpotato.