Abstract: In order to accurately and quickly obtain the water pressure of different parts of the water-passing structure in the hydraulic model test, a method and a system for the intelligent acquisition of the water depth of the piezometric pipe group based on image recognition technology have been proposed and designed. The system adopts a non-contact (shot from a side) image acquisition method which can shoot real-time moving images of water surface buoys in a large number of pressure measuring tubes (reflecting the changing process of water depth), and then import the image data into the computer; the software recognition system adopts grayscale, binary values and opening operations to preprocess the image in preparation. When pretreatment work is finished, findContours would be used to obtain the centroid coordinates of the water surface buoys of each piezometer tube, which would be converted into actual water level data through calibration coefficients, and finally the time series of the synchronous changes of the water depth in each piezometer tube would be obtained. Compared with the traditional manual measurement method, the verification result of hydraulic model test shows that the data error obtained by this method is no more than ±1.5% compared with the traditional manual measurement method, but the acquisition method is faster, the data is richer, and the accuracy is higher.