Abstract: Identifying moving objects from a video sequence is a fundamental and critical task in many computer vision applications, which aims at detecting regions corresponding to moving objects such as vehicles and people in natural scenes. Considering that the existing codebook model algorithm (CBM) can not quite correspond to the computational feature under RGB color space, and does not give simultaneously attention to perturbation resistance and segmentation capability, a fast motion detection method based on improved codebook model is proposed. Pixels are converted from RGB space to YUV space to build the codebook model, which can reduce the computational complexity. After that, the luminance component of each codeword is modeled by the Gaussian model, in order that the codebook model can get the characteristic of the Gaussian mixture model (GMM). So the improved method can combine the advantages of the GMM on the premise of keeping the characteristic of the codebook model. In addition, the method is tested by the typical video sequences, and then the perturbation detection rate (PDR) curves are drawn. Comparative data show that the improved method is more efficient on background segmentation than the CBM algorithm under the RGB space, and can attain a higher capability of anti-perturbation and more adaptively than the traditional CBM and GMM algorithms.