基于红外视觉的点源土壤入渗性能自动测量方法改进

土壤入渗性能指标是农业及其他近地表水文等相关研究领域的关键参数，对其进行快速准确的测量对水资源高效利用具有重要的理论及研究意义。为实现地表有覆盖物时土壤入渗性能的自动精准测量，该研究提出了一种利用热红外成像技术对点源土壤入渗性能测量方法的改进方案，设计并构建了对应的测量系统。采用红外热像仪获取地表湿润面积随时间的推进过程，构建了地表湿润面积提取算法，定义了土壤入渗性能曲线的形状基函数，根据水量平衡原理对曲线配位，并建立了土壤入渗性能模型。室内试验设置了土壤类型（粉壤土和砂壤土）、地表坡度（0°、5°和8°）和覆盖度（20%、40%、50%、60%、65%、70%、75%和80%）3个因素；在自然植被覆盖条件下进行了田间验证试验。结果表明，坡度和土壤类型对地表湿润面积识别影响不显著；当轮廓覆盖度小于60%时，采用本文的算法，地表湿润面积识别误差在5%左右，可满足土壤入渗性能的测量要求。室内及田间入渗性能测量试验的水量平衡误差均在2%以下，表明土壤入渗性能测量模型精度较高。该研究提出的方法操作方便，测量精度高，为实现野外地表有覆盖物的复杂环境下原位测量土壤入渗性能提供了可靠途径。

Method improvement for soil infiltrability measurement based on surface wetted area evolution of the point source infiltration monitored by infrared vision

Soil infiltrability is a one of the key parameters in the fields of agriculture and near surface hydrology. The rapid and accurate measurement has a theoretical and practical significance. However, it is difficult to accurately measure the extremely high infiltration rate in the initial infiltration stage, due to the limitations of measurement principle. In addition, the presence of vegetation cover can also affect the acquisition of surface wetted area in the point source infiltration body, leading to an increase in the measurement error of infiltration rate. In this study, the system was proposed to achieve the automatic and accurate measurement of soil infiltrability with the presence of surface cover. The image correction and preprocessing were adopted to extract the surface wetted area. An infrared thermal imager was also used to record the progress of surface wetted area over time. The shape-based function of soil infiltrability curve was developed using the first-order derivative of the variation of soil wetted area over time. A complete soil infiltrability model was then established, according to the principle of water balance. A device was built to measure the soil infiltrability. The laboratory and field experiments were carried out to validate the measurement accuracy. The indoor experiments were conducted with the soils of sandy loam and silty loam, under surface slope gradients of 0°, 5° and 8°, and surface coverage of 0%, 20%, 40%, 50%, 60%, 65%, 70%, 75% and 80%, respectively. The field validation experiments were conducted on three sites with the natural surface covers, in order to measure the changes in surface wetted area and soil infiltration rate over time. The results show that the surface slope gradient and soil type shared no significant effects on the accuracy of surface wetted area monitoring, whereas, the surface coverage posed a significant impact on the recognition of surface wetted area. Especially, the recognition error of surface wetted area was less than 5% with the surface coverage less than 60%, fully meeting the requirements for soil infiltrability measurement. The measurement error of surface wetted area significantly increased, when the contour line was covered by the vegetation from 70% to 80%. The determination coefficients of the fitted equations between the surface wetted area and time were all greater than 0.98, indicating the better performance of the model on the surface wetted area over time under the vegetation cover conditions. Both surface wetted area function and soil infiltrability function also performed the best in the laboratory and field experiments. The water balance errors of indoor and field infiltration experiments were both less than 2%, indicating a relatively high accuracy of soil infiltrability model without the disturbance of the natural soil surface. The finding can provide an alternative way easy to apply with a sufficient accuracy, particularly for the in-situ measurement of soil infiltrability in filed with the presence of natural surface cover.