Development of the depth measurement system for shallow flow on slopes using an edge-detection algorithm

Abstract: Shallow flow on slopes has been one of the most important driving forces during soil erosion. The simulation of soil erosion can be required to clarify the hydrodynamic conditions and mechanisms in the complex process of water flow and soil. Among them, the water flow depth can be used to calculate the key hydrodynamic parameters for the slope surface flow, such as the water flow power, water flow shear force, and slope surface flow velocity. Therefore, it is very necessary to accurately measure the water depth in the shallow flow on the slopes. However, the slope flow varies greatly in the general open channel, due to a low flow depth, the very complex motion, and hydraulic characteristics. These characteristics make it very difficult to measure the depth of the shallow flow on slopes. In this study, a new depth measurement system was constructed to accurately and efficiently determine the water depth at the position of the shallow flow section on the slope. The horizontal line laser and a high-resolution industrial camera were utilized to capture the pixel coordinate data of the edge of the laser line. An edge detection programming was designed using Python language and OpenCV computer vision library. The functional relationship was then calculated between the offset of the laser line on the surface of the shallow flow and the water depth. The accuracy and stability of the measurement system were verified by the manual measurement, including the probe- and dyeing-based detection. The results show that: 1) The coefficient of variation was only 6.64%±1.40% for the standard water depth measured by this system, indicating higher stability than before. Furthermore, the water depth decreased with the increase of the slope, whereas, that increased with the increase of unit discharge, indicating an excellent consistency and correlation during manual and system measurement. Specifically, the dyeing-based detection presented a closer water depth to the measurement system. Nevertheless, there was a large deviation between the system and manual measurement on the water depth under the unit discharge of 0.333 L/(m·s). The probe-based detection presented an outstandingly larger error than those obtained by the other two methods, where the unit discharge posed a great influence on the measurement accuracy of the probe, whereas, the error of dyeing-based detection mainly came from human observation. Taking the manual measurement as the reference, the Relative Root Mean Square Error (RRMSE), Mean Relative Error (MRE), and Mean Absolute Error (MAE) were all below 0.2, whereas, the Nash-Sutcliffe Efficiency (NSE) reached 0.922 and 0.972, respectively, indicating an excellent consistency and correlation between the water depth system and manual measurement. Consequently, the water depth measurement system can fully meet the high demand to accurately measure the depth of the shallow flow on slopes on a fixed bed, indicating a promising potential to the slope hydrodynamics. 2) The standard deviation was within a reasonable range in the water depth measurement system at the low fluctuation degree of the water flow. Once the standard deviation in the water depth measurement system was more than the manual, the flow area to which it belongs can be a rolling wave flow area, indicating a higher accuracy of the system. More importantly, the water depth of the entire shallow flow section can also be collected frame by frame, further to respond and record the violent fluctuations and dynamic changes of the waveform generated by the water surface.