基于聚类与链码技术的交通标志检测

Traffic Sign Detection Based on Clustering and Chain Code Technique

交通标志的有效检测是交通标志识别系统中的关键步骤;提出一种基于颜色和形状的交通标志检测新方法,首先由最小欧式距离的聚类分析方法以及特征量与聚类中心的向量积提取夹角正弦方法,构造两级颜色特征分类器并通过训练实现优化分割,然后对滤波后图像边界跟踪,利用新的链码方法实现区域的拐角点提取,最后由几何特征判定区域的形状进行定位,实现交通标志的检测;实验结果表明,该方法在不同气候条件下的平均检测率达92.96%,优于同类方法且具有较高的鲁棒性。

Detecting traffic sign effectively is a key technique in traffic sign recognition system. A method of traffic sign detection based on color and shape was presented. First, the method of clustering analysis of minimum Euclidian distance was used to construct the first color classifier. The sine value was derived fron~ the cross product of the characteristic parameters and the cluster center, and it was presented for constructing the second color classifier. The segmentation of traffic sign is achieved by training the classifier. And then, the algorithm of edge tracing was used to the filtered image. A new method of chain code was presented to extract the corner points of the region. Finally, the shape of the region was identified by the geometric feature, and the detection results were improved by traffic sign position. Experimental result shows that the detection rate is 92. 96% in different climatic conditions; the method outperforms the previously existing method and has better robustness.