APPLICATION OF INTEGRAL LENGTH SCALE AND CONVOLUTIONAL NEURAL NETWORKS FOR HYDROLOGICAL MEASUREMENT
Part Of
Proceedings of the IAHR World Congress
Start Page
2391
End Page
2394
ISSN
25217119
ISBN (of the container)
9789083558950
ISBN
9789083558950
Date Issued
2025
Author(s)
Abstract
Discharge is a key parameter in hydraulic engineering, particularly for structural design, disaster prevention, and water resource management. Accurate discharge data is crucial for effective decision-making and planning. However, conventional measurement techniques, such as Acoustic Doppler Current Profilers (ADCP), present several disadvantages, including high costs, time consumption, and safety risks to personnel. Additionally, radar and ultrasonic-based measurement techniques, while widely used in field applications, are highly susceptible to environmental factors and provide only single-point data. These limitations highlight the necessity for innovative methods to enhance the efficiency of hydrological measurement. In response to these challenges, this study applied a novel velocity measurement method based on the Large-scale Particle Image Velocimetry (LSPIV) combined with Convolutional Neural Network (CNN) algorithms. This method can significantly reduce the impact of environmental noise and improve accuracy. The velocity data can also be used to calculate the integral length scale, which is defined as the integral of the normalized spatial autocorrelation function of turbulent velocity fluctuations. By analyzing the relationship between the integral length scale and water depth, we can invert two-dimensional water depth and estimate discharge. To validate the effectiveness of this method, the experiments were conducted in a controlled flume equipped with longitudinal and transverse bed structures, as well as in-field river channels. In field experiments, images were captured by a drone and stabilized using Scale Invariant Feature Transform (SIFT) techniques. The results demonstrate that this two-dimensional, non-contact measurement technique not only effectively estimates discharge but also offers advantages such as low operational costs, high efficiency, and enhanced safety for personnel. Therefore, this research has the potential to emerge as a novel hydrological measurement method.
Event(s)
Book of Extended Abstracts of the 41st IAHR World Congress, 2025, Singapore, 22 June 2025 - 27 June 2025
Subjects
Convolutional Neural Network
Hydrological measurement
Integral length scale
LSPIV
Publisher
International Association for Hydro-Environment Engineering and Research
Type
conference paper