OD matrix is one of the most important research fields in intelligent traffic, it's the key data of microscopic traffic simulation, city traffic planning, management and control etc…In this article, through the multipath search algorithm based on point and punishment method and modified Logit model to determine the distribution of matrix, according to the structure of OD deviation state space model and combining with polynomial trend based on information fusion Kalman filtering, put forward the dynamic OD matrix estimation algorithm based on information extraction. Finally, combining with the example of analysis and calculation, verify the validity and feasibility of the presented method in this paper. Keywords: Information extraction technologies; Dynamic OD matrix estimation; Distribution matrix; Kalman filter.