Abstract: A method was proposed to build background model of passenger flow image in comprehensive passenger transport hub based on dempster-shafer. Mean background model and gray area division were used to improve processing speed of background modeling. Mass function which was from Dempster-Shafer was used to improve confidence of background gray value area and accuracy of background image. Test results show that principle of the method is correct; passenger flow background image can be quickly and accurately built. This method insured the speed and precision of passenger flow information extraction in the passenger flow safety forewarning of high-speed railway comprehensive passenger transport hub.