The rise of agricultural Internet of Things (IoT) technologies is transforming traditional farming into modern, sustainable agriculture through scientific advancements. This study introduces a novel fusion calibration algorithm that combines the least squares method and back propagation neural networks to enhance water measurement accuracy in agricultural settings. By developing a nonlinear function, the algorithm progressively minimizes the discrepancies between detected results and actual data until satisfactory accuracy is achieved. An irrigation experiment on wolfberry plants in City H utilizing this optimized technology demonstrated a near-perfect correlation between the automatic measurements (3.56 m/s) and the actual flow rates (3.55 m/s) recorded by a flow meter, with an error margin of just 0.282%. Furthermore, the study observed a steady increase in the water utilization coefficient in farmland irrigation from 0.54 in 2011 to 0.589 in 2020, indicating enhanced water efficiency and conservation.