中文摘要：

用GA、PCA和改进SVM结合进行车辆实时分类研究对公路管理和控制具有实际应用意义和社会效益。在某公路的匝道口两侧设置了8个测试点,对通过匝道口的车辆进行测试,提取特征向量,采用声波和振地波信号在匝道口进行了实时分类测试研究。由于特征向量的维数太高,用GA和PCA降低特征向量的维数,再用SVM和改进的SVM对特征向量进行分类,大大提高了分类精度。通过实验及分类得到声波和振地波的测试集分类精度最高分别是92.0%和76.1%,同时声波和振地波特征向量的维数降低至26和21,其相应的比率分别为95%和99%,独立集的分类精度分别为87.5%和71.3%。实验表明：用改进的SVM和GA、PCA结合的方法进行分类,其效果要优于单独使用主成分分析（PCA）、遗传算法（GA）以及它俩的结合使用的方法。

英文摘要：

It is practically and socially significant to real-time classificate vehicles using GA,PCA and improved SVM.Eight test points were set up on both sides of a road ramp to collect feature vectors of passing vehicles.Real-time classification test was conducted with acoustic and seismic signals.Over-high dimension of feature vectors was reduced using GA and PCA.Then,the feature vectors were classified with SVM and improved SVM;thereby the classification accuracy was greatly improved.The highest classification accuracy of acoustic and seismic signals obtained by experiments was 92.0% and 76.1%.The dimension of feature vectors of acoustic and seismic signals was reduced to 26 and 21,respectively,with corresponding ratio of 95% and 99%.The corresponding classification accuracy of independent set was 87.5% and 71.3%.Experiment result shows that the classification accuracy using GA,PCA and improved SVM is much higher than that of using PCA,GA alone or using two of them.