Abstract:The robot in the RoboCup mid-size group match must be capable of identifying the positions and directions of other objects in the match through predefined color patterns in real time. Because of the existence of noise and the complexity of signal objects, the identification algorithm must assure the whole process of identification stable and make the response time as short as possible. The objects should be correctly classified and the possible exceptional conditions must be taken care of. So one has to study various kinds of misidentification. This paper proposes a possible algorithm to deal with this problem. The algorithm minimizes the effects of noise through the use of the prediction of the objects' moving direction and the characteristic information of objects like the color pattern, area change ratio and the density of color dots. The experiment confirms that the algorithm can minimize the effects of noise, identify objects fast and accurately under the conditions of a match.