无人机热红外图像计算冠层温度特征数诊断棉花水分胁迫

针对当前无人机热红外遥感诊断作物水分胁迫状况精度不高的问题,该文以4种水分处理的花铃期棉花为试验对象,利用六旋翼无人机搭载热红外传感器,连续5 d采集中午13点的棉花冠层高分辨率热红外影像,通过Canny边缘检测算法将热红外图像中的土壤背景有效剔除,应用温度直方图验证剔除效果,然后计算棉花冠层温度特征数,包括冠层温度标准差(standard deviation of canopy temperature,CTSD)和冠层温度变异系数(canopy temperature coefficient of variation,CTCV);分别研究棉花冠层温度特征数与棉花叶片气孔导度Gs、蒸腾速率Tr、水分胁迫指数(crop water stress index,CWSI)和土壤体积含水率(soil volumetric water content,SWC)的相关关系,并分析冠层温度特征数对诊断棉花水分胁迫的适用性。研究结果表明:棉花冠层温度特征数与表征棉花水分胁迫的生理指标和物理指标都具有较高的相关性,最大的决定系数R2为0.884;棉花冠层温度标准差CTSD和变异系数CTCV与Gs、Tr、CWSI、SWC的决定系数R2分别为0.884、0.625、0.673、0.550和0.853、0.583、0.620、0.520,冠层温度标准差CTSD对作物水分胁迫的敏感程度更高,可以作为诊断作物水分胁迫的新指标。该研究提出冠层温度特征数的计算方法仅需要无人机热红外影像数据,相比其他诊断作物水分胁迫状况的温度指标具有较大的应用潜力。

Cotton moisture stress diagnosis based on canopy temperature characteristics calculated from UAV thermal infrared image

Abstract: The canopy temperature can indicate crop water stress condition, but the traditional method of canopy temperature measurement is based on ground point measurement, time-consuming and laborious. The calculation of crop water stress index (CWSI) requires more ground data monitoring and the crop water deficit physiological indicators are also difficult to measure in specific practical application, so it is difficult to popularize widely in agriculture. With the rapid development of unmanned aerial vehicle (UAV) platforms and thermal infrared sensors, the thermal infrared technology of UAV can monitor crop canopy temperature quickly and effectively and obtain a large area of crop canopy temperature with fast, dynamic monitoring, all-weather operation advantages, and further diagnose the water stress condition of agricultural crops. To solve the problem of low precision of the current UAV thermal infrared remote sensing for the diagnosis of crop water stress, in this paper, the 4 water treatment plots I1, I2, I3 and I4 with 50%, 65%, 80% and 95%-100% of field holding water as the upper limit were set up, each treatment plot had 3 repeated tests, a total of 12 treatment plots, and the cotton at flower boll period under 4 kinds of water treatments was selected as the test object. A six-rotor unmanned aerial vehicle was used to carry the thermal infrared sensor, and the high resolution thermal infrared images of cotton canopy at 13 o''clock at noon were collected in 5 consecutive days. First, the custom coordinates for thermal infrared images are defined in order to make the image have the same operating position, and Canny edge detection algorithm is used to obtain the edge feature raster image of cotton canopy. Then, the edge feature raster images are processed in ArcGIS and ENVI (environment for visualizing images) software to obtain a polygon vector layer, including the cotton canopy edge feature raster image converting to polyline vector layers and the polyline vector layers converting to polygon vector layers. Finally, use the polygon vector layer to clip, mask statistics and draw the canopy temperature histogram. Through the Canny edge detection algorithm, the thermal infrared image of the soil background is effectively removed, and the application of canopy temperature histogram verifies the elimination effect. The number, position and size of the original temperature histogram''s peaks correspond to different matter pixels. If the temperature histogram is a single peak histogram for the high coverage crop canopy, the crop canopy with low coverage will be characterized by double or triple peaks. According to the canopy temperature histogram, the temperature characteristics of cotton canopy are calculated, including standard deviation of canopy temperature (CTSD) and canopy temperature coefficient of variation (CTCV). The relationships between cotton canopy temperature characteristics and cotton leaf stomatal conductance (Gs), transpiration rate (Tr), crop water stress index (CWSI) and soil water content (SWC) were studied, and the applicability of canopy temperature characteristics to diagnose cotton water stress was analyzed. The results showed that the cotton canopy temperature characteristics were correlated with the physiological and physical indices representing cotton water stress, and the maximum determining coefficient (R2) was 0.88. The determining coefficients of CTSD and CTCV with Gs, Tr, CWSI and SWC were 0.884, 0.625, 0.673, 0.550 and 0.853, 0.583, 0.620, 0.520, respectively. The CTSD was more sensitive to crop water stress condition, which could be used as a new index to diagnose crop water stress. However, the correlation between CTCV and physiological and physical indicators of crop water deficit is not ideal. Compared with CWSI, Gs, Tr and SWC which are traditional diagnostic indicators of crop water stress, this method greatly simplifies the calculation of crop diagnostic indicators. In this research, it suggests that the calculation method of canopy temperature characteristics only needs the thermal infrared image data of UAV and does not require any meteorological factors, so the canopy temperature characteristics are more easily promoted for use in agricultural water-saving irrigation. It has great potential to be used in the diagnosis of crop water stress compared with other water stress indices.