节点文献
基于Bayesian-XGBoost的生菜作物系数估算方法
Estimation Method of Lettuce Crop Coefficient Based on Bayesian-XGBoost
【摘要】 准确估算作物系数(K_c)对于计算作物蒸发蒸腾量(ET_c),实现精确有效的用水管理至关重要。为了及时掌握K_c的连续动态变化,进一步估算作物需水量来指导作物灌溉,采用超绿算法和最大类间方差法进行图像分割来提取生菜冠层覆盖度(PGC),并提出了基于贝叶斯优化XGBoost(Bayesian-XGBoost)的生菜K_c估算模型,为指导作物科学合理灌溉提供新途径。结果表明,全生育期内PGC呈先快速增长后趋于稳定的生长趋势,变化范围为10.38%~81.00%,日均增幅为1.56%,可有效反映K_c的变化趋势。以PGC作为输入所构建的BayesianXGBoost生菜K_c估算模型在迭代400次时达到最优估算效果,生菜幼苗期、莲座期、结球期的K_c分别为1.12、1.29和2.26,与实测值相比平均误差为2.13%。此外,进一步利用Penman-Monteith公式(PM公式)结合K_c估算模型实现了对生菜ET_c的实时计算,得到了秋茬种植期的生菜日均ET_c为2.38 mm/d。基于Bayesian-XGBoost的生菜K_c估算模型能够较好估算生菜K_c和作物需水量。
【Abstract】 Accurate estimation of crop coefficient(K_c) is crucial for calculating crop evapotranspiration(ET_c), and achieving accurate and effective water management. In order to grasp the continuous dynamic change of K_c in time and further estimate the crop water requirement to guide crop irrigation,in this paper, the super-green algorithm and maximum interclass variance method was adopted for image segmentation to extract the lettuce canopy coverage(PGC), and a lettuce K_c estimation model based on Bayesian optimization XGBoost(Bayesian-XGBoost) was proposed. It provided a new way to guide scientific and rational irrigation for crops. The experimental results showed that: PGC showed a rapid growth trend at first and then tended to be stable during the whole growth period, with a range of 10.38%-81.00% and an average daily increase rate of PGC of 1.56%, which could effectively reflect the change trend of K_c. The Bayesian XGBoost lettuce K_c estimation model constructed with PGC as input achieved the optimal estimation effect when iterating for 400 times. K_c of lettuce at seedling stage, rosette stage, and heading stage was 1.12, 1.29, and 2.26, respectively, with an average error of 2.13% compared with the measured true value.In addition, penman-monteith formula(PM) was further used with K_c estimation model to realize the real-time calculation of lettuce ET_c. The average daily ET_c of lettuce in the autumn cropping period was 2.38 mm/d. The lettuce K_c estimation model based on Bayesian XGBoost could better estimate lettuce K_c and crop water requirement.
【Key words】 potted lettuce; crop coefficient; image processing; Bayesian optimization; machine learning; crop evapotranspiration;
- 【文献出处】 山西农业科学 ,Journal of Shanxi Agricultural Sciences , 编辑部邮箱 ,2022年10期
- 【分类号】S636.2
- 【下载频次】117