Rational utilization of water resources is one of the major methods of water conservation. There are significant differences in the irrigation needs of different agricultural fields because of their spatial variability. Therefore, a decision support system for variable rate irrigation (DSS-VRI) by center pivot was developed. This system can process multi-spectral images taken by unmanned aerial vehicles (UAVs) and obtain the vegetation index (VI). The crop evapotranspiration model (ET_c) and crop water stress index (CWSI) were obtained from their established relationships with the VIs. The inputs to the fuzzy inference system were constituted with ET_c, CWSI and precipitation. To provide guidance for users, the duty-cycle control map was outputted using ambiguity resolution. The control command contained in the map adjusted the duty cycle of the solenoid valve, and then changed the irrigation amount. A water stress experiment was designed to verify the rationality of the DSS-VRI. The results showed that the more severe water stress is, the more irrigation is obtained, consistent with the expected results. Meanwhile, a user-friendly software interface was developed to implement the DSS-VRI function.

As the key principle of precision farming, variation of actual crop evapotranspiration (ET) within the field serves as the basis for crop management. Although the estimation of evapotranspiration has achieved great progress through the combination of different remote sensing data and the FAO-56 crop coefficient (Kc) method, lack of the accurate crop water stress coefficient (Ks) at different space–time scales still hinder its operational application to farmer practices. This work aims to explore the potential of multispectral images taken from unmanned aerial vehicles (UAVs) for estimating the temporal and spatial variability of Ks under the water stress condition and mapping the variability of field maize ET combined with the FAO-56 Kc model. To search for an optimal estimation method, the performance of several models was compared including models based on Ks either derived from the crop water stress index (CWSI) or calculated by the canopy temperature ratio (T_{c ratio}), and combined with the basal crop coefficient (Kcb) based on the normalized difference vegetation index (NDVI). Compared with the Ks derived from the T_{c ratio}, the CWSI-based Ks responded well to water stress and had strong applicability and convenience. The results of the comparison show that ET derived from the Ks-CWSI had a higher correlation with the modified FAO-56 method, with an R² = 0.81, root mean square error (RMSE) = 0.95 mm/d, and d = 0.94. In contrast, ET derived from the Ks-T_{c ratio} had a relatively lower correlation with an R² = 0.68 and RMSE = 1.25 mm/d. To obtain the evapotranspiration status of the whole maize field and formulate reasonable irrigation schedules, the CWSI obtained by a handheld infrared thermometer was inverted by the renormalized difference vegetation index (RDVI) and the transformed chlorophyll absorption in reflectance index (TCARI). Then, the whole map of Ks can be derived from the VIs by the relationship between CWSI and Ks and can be taken as the basic input for ET estimation at the field scale. The final ET results based on multispectral UAV interpolation measurements can well reflect the crop ET status under different irrigation levels, and greatly help to improve irrigation scheduling through more precise management of deficit irrigation.

An improved semantic segmentation method based on object contextual representations network (OCRNet) is proposed to accurately identify zucchinis intercropped with sunflowers from unmanned aerial vehicle (UAV) visible images taken over Hetao Irrigation District, Inner Mongolia, China. The proposed method improves on the performance of OCRNet in two respects. First, based on the object region context extraction structure of the OCRNet, a branch that uses the channel attention module was added in parallel to rationally use channel feature maps with different weights and reduce the noise of invalid channel features. Secondly, Lovász-Softmax loss was introduced to improve the accuracy of the object region representation in the OCRNet and optimize the final segmentation result at the object level. We compared the proposed method with extant advanced semantic segmentation methods (PSPNet, DeepLabV3+, DNLNet, and OCRNet) in two test areas to test its effectiveness. The results showed that the proposed method achieved the best semantic segmentation effect in the two test areas. More specifically, our method performed better in processing image details, segmenting field edges, and identifying intercropping fields. The proposed method has significant advantages for crop classification and intercropping recognition based on UAV visible images, and these advantages are more substantive in object-level evaluation metrics (mIoU and intercropping IoU).