Abstract: A recognition method based on the machine vision was designed and investigated to solve the problem on the measurement of molten steel level in the strip casting processing. Based on the different image grey values captured by linear CCD camera between the molten steel and casting rolls， a machine vision system for measuring molten steel level was designed. By the high speed character of the linear camera image data acquisition， a method to create molten steel level original image by the time was proposed and the molten steel level recognition model was established building with filtering， binarization and boundary recognition as the core. The system and model were tested on a smallscale strip casting machine and the induction furnace respectively. The results illustrate that the molten steel level detection system and its recognition method are feasible， which show high levels of the precision and response speed and can be applied in the industrial line for the strip casting.

Key words: molten steel level, CCD camera, image treatment, edge recognition, twinroll strip casting