In order to solve the problem of false detection of the moving liquid level caused by the vibration of the constant temperature water bath, this paper combines the Type-2 Fuzzy Gaussian Mixture Model (T2-FGMM) and Markov Random Field (MRF) to study a new background modeling method for detecting the moving liquid level in dynamic scenes. The method first considers the output of T2-FGMM as the initial labeling domain of MRF, and then combines the local energy of the labeling domain with the observation energy. The key of this method is to combine the spatiotemporal prior of T2-FGMM with the observation. Comparative experimental results show that the proposed algorithm has better dynamic background detection effect than traditional frame difference method and Vibe algorithm, and can effectively eliminate the influence of the vibration of the constant temperature water bath on the detection of the moving liquid level of the viscometer.