水胁迫下多角度幼龄檀香图像颜色变化分析及含水率反演

干旱与水淹胁迫是植物遭受的主要非生物胁迫,对植物的生理活动造成严重影响.本研究基于单反相机获取幼龄檀香的纵向和冠层叶片图像,使用分割算法提取叶片和颜色特征,然后讨论两种胁迫条件下多角度檀香叶片颜色变化及含水率反演.结果表明:干旱组在胁迫前期(前6d)叶片亮度降低,绿色分量增加,之后叶片亮度增加,绿色分量降低;水淹组叶片在整个胁迫周期亮度持续降低,黄色分量增加;对照组则与干旱组的变化趋势类似,但拐点出现的时间较晚.当叶片含水率在50%~70%时,随着含水率的增加,彩色图像的红(R)、绿(G)、蓝(B)通道值均会减小;但当叶片含水率小于40%时,会出现R通道值大于G通道值的现象.在使用极限学习机反演含水率时,校正后的颜色分量对拟合优度及预测精度均有所提高.纵向图像更适合用来反演叶片的含水量,决定系数和平均绝对百分比误差分别为0.8352和2.3%;而冠层图像对叶片等效水厚度的表达更准确,上述指标分别为0.7924和9.3%.

Color change analysis and water content inversion of young sandalwood in multi-angle under water stress

Drought and waterlogging are two main abiotic stresses for plants, with serious impacts on plant physiological activities. In this study, the vertical and canopy leaf images of young sandalwood were obtained by SLR camera, with leaf segmentation algorithm being used to extract leaves and color features. We examined the color change of sandalwood leaves and water content inversion in different angles under two stress conditions. The results showed that leaf brightness decreased while the green component increased in the early stage (the first six days) of drought stress. After that, the brightness began to increase and green component began to decrease. Under water stress, the brightness of leaves decreased and yellow component increased in the whole stress cycle. The changes of control group was similar to that of the drought group, but the inflection point appeared later. Under the range of 50% to 70% for water content of leaves, the value of R, G, B channel of color images would decrease with the increases of water content. When the water content of leaves was less than 40%, the R channel value was larger than the G channel value. When using the extreme learning machine to retrieve the water content index, the corrected color components improved the fitness and the prediction accuracy. The vertical image was more suitable for retrieving water content of leaves, with the error of determination coefficient and average absolute percentage being 0.8352 and 2.3%, respectively. The canopy images were more accurate in expressing the equivalent water thickness of blades, with the above indices of 0.7924 and 9.3%, respectively.