Abstract: Tags anti-collision algorithm is the key to the efficiency of radio frequency identification system.At present the main improvement on framed slotted Aloha algorithm is to adjust the frame size in the process of identification，but there are such disadvantages as inaccurate calculating frame size，complex algorithm and long identification time.This paper proposes a new anti-collision algorithm that divides the identification period into two steps，which are collision detection and data reading.It makes full use of collision information of the tag detection to increase the efficiency of data reading，reduce the identification time and improve the identification rate.Theoretical analysis and simulations show that the algorithm can identify the tags more efficiently and rapidly，especially in the case of multi-target recognition.