Improved moving objects detection method based on codebook model

摘要

Traditional Gaussian mixture model cannot cope with illumination changes in the scene and slowly-moving objects. Codebook model cannot detect foreground objects in the training process. The algorithm is proposed in the paper. It uses the background of Gaussian mixture model to build codebook model. Using Frame difference to determine changes in the environment and update the codebook model. This approach solves the problem of mutant changes in the environment. Experiments show that the approach can detect foreground object in complex environment, eliminate shadows effectively and detect the foreground object more accurately.