The leaf water content of plants responds to the growth of the plant. In this paper, a method of nondestructive detection of the leaf water content by image analysis was proposed. Taking the cucumber leaves as the research object, the whole plant images were collected in batches of cucumber plants at different developmental stages. The target leaf images were obtained by GrabCut algorithm, and the color, gray statistics and texture features were calculated. Then, a regression model with the leaf water content was established. The results showed that the established regression model with R2 kept good predictive ability for the water content of cucumber leaves at different growth stages. The average relative errors between the two batches of test data were 10. 88％ and 7. 98％ . The model could be used for nondestructive testing of cucumber leaf water content and could be combined with high-throughput phenotyping platform to improve the accuracy of continuous monitoring of cucumber leaf water content.%植物叶片的含水量能反映植物的生长状况.为了实现通过图像分析技术对植物叶片的含水量进行无损检测,以黄瓜叶片为研究对象,针对不同发育时期的黄瓜植株, 分批次采集植物整株图像.使用 GrabCut算法分割得到目标叶片图像,融合计算灰度统计特征及纹理特征,与叶片含水量建立回归模型.结果表明: 建立的回归模型 R2 为 0. 8358,对于不同生育期的黄瓜叶片的含水量均有较好的预测能力,两批测试数据的平均相对误差为 10. 88％ 和 7. 98％ .该模型可以用于黄瓜叶片含水量无损检测,可与高通量表型平台相结合,提高黄瓜叶片含水量变化连续监测的精确度.
展开▼
