This paper presents a new method for irregular shape recognition based on traffic sign recognition. The projection sequence is obtained by the bidirectional projection transformation of the irregular edge points. The weighted mass is obtained according to the normalized and centered projection sequence data, and the centroid of the projection sequence is obtained. Because the centroid distance of the similar particle system is similar, the centroid of the similar projection sequence is similar. By using the linear combination of the chessboard distance and the block distance to replace the Euclidean distance to measure the similarity of the projection sequence and its centroid, and then get the similarity degree of the irregular shape. The experimental results show that the algorithm has high recognition accuracy and fast computation speed.