The potential of combining artificial neural networks (ANNs) and image processing for assessing leaf relative water content (RWC) and water content (WC) was addressed. Spathiphyllum wallisii was employed as model species, because it has broad leaves and very responsive stomata. In the course of desiccation, leaves were periodically weighted (to calculate RWC and WC conventionally) and imaged. Image acquisition was performed by a scanner, and was, thus, independent of ambient light environment. Color feature extraction was performed in three color spaces (RGB, HSI, and CIELAB), while six texture statistical features were calculated for each of the (nine) computed color channels. Prior to model development via ANNs, the obtained feature vector underwent feature reduction using principal component analysis. The presented methodology yielded very precise estimations of leaf RWC and WC (correlation coefficient > 0.95). Therefore, the technique under study was proven to be very promising for non-invasive in situ measurements of leaf water status.

提出了结合人工神经网络（ANN）和图像处理来评估叶片相对含水量（RWC）和含水量（WC）的潜力。桔梗它被用作模型物种，因为它具有宽阔的叶子和非常灵敏的气孔。在干燥过程中，定期对叶子进行加权（以常规方式计算RWC和WC）并成像。图像采集是由扫描仪执行的，因此与环境光环境无关。在三个颜色空间（RGB，HSI和CIELAB）中执行颜色特征提取，同时为（九个）计算出的每个颜色通道计算了六个纹理统计特征。在通过人工神经网络进行模型开发之前，使用主成分分析对获得的特征向量进行特征约简。提出的方法得出了叶片RWC和WC的非常精确的估计值（相关系数> 0.95）。所以，