Supplementary data for RivQNet: Deep Learning based river discharge estimation using close-range water surface imagery

The supplementary data for the "RivQNet: Deep Learning based river discharge estimation using close-range water surface imagery", recorded footage and measured ADCP data used for validation.
In this study, we present RivQNet, a novel image-based method for measuring streamflow that utilizes artificial intelligence techniques and does not require subjective user input. RivQNet processes close-range, non-contact images of the water surface using a convolutional neural network architecture called FlowNet. The accuracy of RivQNet is validated through comparison with standard measurement methods and conventional optical flow methodologies. The results show that RivQNet produces accurate and dense spatial distributions of surface velocities. Streamflow data is an essential input for many hydrological and hydraulic research, modeling, and design studies, but current image-based surface velocimetry techniques that use correlation approaches can be biased if the operator is not experienced. Our goal in this study was to develop a more accurate and efficient river velocimetry method that does not rely on subjective user input.

本数据集为“RivQNet：基于深度学习的近程水面影像河流流量估算”的补充数据，其中包含用于验证的记录影像和测量到的 ADCP 数据。在本研究中，我们提出了 RivQNet，这是一种新型的基于图像的河流流量测量方法，它运用人工智能技术，无需依赖主观的用户输入。RivQNet 通过 FlowNet（一种卷积神经网络架构）处理近程、非接触式的水面图像。通过将 RivQNet 的精度与标准测量方法和传统光学流方法进行比较，验证了其准确性。结果显示，RivQNet 能够生成精确且密集的表面速度空间分布。河流流量数据对于众多水文和水力研究、建模及设计研究至关重要，然而，目前基于图像的表面流速测量技术，若操作者缺乏经验，使用相关性方法可能会产生偏差。本研究旨在开发一种更精确、更高效的河流流速测量方法，该方法不依赖于主观用户输入。