The present study provides an assessment of the genetic variability of different genotypes of Passiflora using physical and chemical descriptors and digital image analysis to quantify leaf, flower and fruit coloration as well as to develop and validate mathematical models for the prediction of passion fruit pulp weight without seeds by regression analysis. A total of 132 Passiflora genotypes belonging to different species were evaluated. Eleven quantitative physical and chemical descriptors were submitted to principal component and cluster analysis. The coloration was determined through RGB color space by digital images analysis. The equations were developed through multiple linear regression and cross-validation was used to validate the models. The first principal component (50.6% of total variance) was associated with fruit weight, diameter and length and peel weight, while the second (22.4%) was associated with fruit weight, fruit length, peel weight, peel thickness, soluble solids, fruit length/fruit diameter ratio and fruit length and diameter. Physical and chemical descriptors showed significant diversity among genotypes. In the cluster analysis, the genotypes BGP007 and BGP009 were most closely related, while BGP349 and BGP177 were distinct. The analysis of coloration using digital images allowed detecting greater variability in the color of leaves, flowers, fruits, and pulp. The analysis of digital images is a practical method for ascertaining the color of plant parts in the field or laboratory. The high phenotypic diversity detected in the present study can be used in passion fruit breeding programs through hybridizations. Regression analysis of physical traits revealed five models that could be used to estimate the pulp weight in two equations (13 and 15), which showed the highest R² values and lowest RMSE and MAE. This model can be reliably adopted for estimation of the pulp weight without seeds of passion fruits (P. edulis Sims).

本研究使用物理和化学描述符和数字图像分析对西番莲不同基因型的遗传变异进行评估，以量化叶、花和果实的颜色，并开发和验证用于预测西番莲果肉重量的数学模型，无需通过回归分析种子。一共132个西番莲评估了属于不同物种的基因型。将 11 个定量物理和化学描述符提交给主成分和聚类分析。通过数字图像分析通过 RGB 颜色空间确定着色。这些方程是通过多元线性回归开发的，并使用交叉验证来验证模型。第一个主成分（总方差的 50.6%）与果实重量、直径和长度以及果皮重量相关，而第二个（22.4%）与果实重量、果实长度、果皮重量、果皮厚度、可溶性固形物、果实果长/果径比和果长和果径。物理和化学描述符显示基因型之间的显着差异。在聚类分析中，基因型 BGP007 和 BGP009 的关系最密切，而 BGP349 和 BGP177 是不同的。使用数字图像进行着色分析可以检测叶子、花朵、果实和果肉颜色的更大变化。数字图像分析是在田间或实验室中确定植物部分颜色的实用方法。本研究中检测到的高表型多样性可通过杂交用于百香果育种计划。物理特性的回归分析揭示了可用于估计两个方程（13 和 15）中纸浆重量的五个模型，其中显示最高 数字图像分析是在田间或实验室中确定植物部分颜色的实用方法。本研究中检测到的高表型多样性可通过杂交用于百香果育种计划。物理特性的回归分析揭示了可用于估计两个方程（13 和 15）中纸浆重量的五个模型，其中显示最高 数字图像分析是在田间或实验室中确定植物部分颜色的实用方法。本研究中检测到的高表型多样性可通过杂交用于百香果育种计划。物理特性的回归分析揭示了可用于估计两个方程（13 和 15）中纸浆重量的五个模型，其中显示最高R ²值和最低的 RMSE 和 MAE。该模型可以可靠地用于估计没有百香果种子（P. edulis Sims）的果肉重量。