The moving objectives detecting technology in video surveillance system was stu-died. On the base of analysis and comparison of several target detection algorithm

a way of using block processing method to establish the initial background model was proposed. The method had a good inhibitory effect on the noise in video sequences

and it can overcome the shortcomings of fuzzy and residual shadow caused by mean method; In order to meet the change of background

the three-frames differences method was used to find those pixels that had a significant difference

then these pixels were updated adaptively. Simulation experiment showed that

this method can be used in background subtraction algorithm. The moving targets can be detected rapidly

completely and accurately.