Abstract: A head sheave deflection detection system based on machine vision was designed. First, CCD camera of the machine vision system is used to have the real-time acquisition of the running state video of the head sheave deflection; then, each frame of image is carried on the filtering and enhancement of the image, and Mean Shift algorithm is used to realize the tracking of the head sheave deflection; finally, the pixel offset of the track target is converted for the actual physical dimensions to calculate the head sheave deflection through the coordinate transformation. The experimental results show that the field detection and installation of the designed deflection sheave detection system is convenient and can meet the requirements of precision with good practical value. It solves the problem that most mine friction head sheave deflection has difficult detection and high cost.