Abstract: Breakthroughs have been made in artificial intelligence technologies including reasoning，multimodal
processing，intelligent agents，image recognition and object detection．These advances will profoundly transform
the content and implementation of physical experiments，making the integration of AI into physical experiments
an imperative．The YOLO series models，a key branch of AI technology，are widely applied and deliver excellent
performance in image recognition and object detection．In the falling-ball method for measuring the viscosity
coefficient of liquid，observations and image recording traditionally rely entirely on manual operation．Major
challenges include objectively judging whether the ball falls vertically，eliminating parallax errors and reducing
timing deviations． By introducing YOLOv5 into the conventional falling-ball experiment for liquid viscosity
coefficient measurement，we realize automatic recognition，trajectory tracking and velocity calculation of the
falling ball．Images are captured in real time by a high-definition industrial camera．Combined with the pre-
trained YOLOv5model，the system can accurately extract the position data of the ball and calculate the viscosity
coefficient based on Stokes' law． Experimental results demonstrate that this system greatly improves
measurement accuracy and efficiency．It can also identify the deflection of the falling trajectory． This study
provides a feasible approach for the interdisciplinary integration of physical experiments and artificial
intelligence，and boasts great potential for promotion in teaching and application in scientific research．
