基于双向二维主成分分析的交通标志识别

Traffic Sign Recognition Based on Two-directional 2D Principal Component Analysis

针对交通标志识别实时性不足,提出了一种基于双向二维的主成分分析[(2D)2PCA]的交通标志识别算法。首先,对交通标志图像进行去噪归一化等预处理。然后,进行水平和垂直方向的投影,通过特征空间降维提高匹配速度。最后,利用最近邻法进行分类。通过在不同数据库下与传统2DPCA方法的对比仿真表明,2种方法随主特征数目增加,识别率都有所提升;样本数量增加时,(2D)2PCA算法的时间增长速度明显小于2DPCA,满足了识别的实时性要求。

A traffic sign recognition algorithm based on two-directional two-dimensional principal component analysis [ （2D）2 PCA ] method is proposed to improve real-time recognition. In this approach, the traffic sign images are firstly denoised and normalized in the preprocessing stage. Then, these images are projected onto horizontal and vertical axes, which led to dimensionality space reduction suitable for fast matching. Finally, the nearest neighbor method is involved for classification on these dimensionality-reduced images. Comparison with simulation result of traditional 2DPCA based on different datasets show that （1） both methods can improve their recognition rates along with an increasing number of PCs; （2） （2D）2PCA＇s algorithm time growth rate is significantly less than that of 2DPCA with an increasing number of samples, which verified the ability of the proposed method on real-time recognition task.