基于最优参数搜索的车辆中网识别方法研究

目前国内外车型识别方法中基于中网区域特征的研究较少，且分类识别的效率和精度较低。该文在分析中网格栅区域结构特征、中网窗口形状特征及区域纹理特征的基础上，提出基于最优参数搜索的改进型C参数的支持向量分类(C-SVC)车辆中网分类识别方法，该方法采用双角度约束以提高分类的效率和精度，即一方面设计基于马氏距离和a-原则对样本数据进行优化分选，并结合加权判别算法加快支持向量机的训练测试速度，以提高算法泛化效率；另一方面在核函数参数设定过程中，设计了基于先验知识的迭代最优参数搜索算法，以提高分类器的分类识别精度。实验表明，上述车辆中网识别方法检测准确率达到97.53%，具有精度高、误检率低的优点，同时极大优化分类识别效率，能够满足识别分类的实时性要求。

Study on Vehicle Grille Recognition Method Based on the Optimal Parameter Searching

There are few studies on the vehicle recognition methods based on grille regional characteristics both at home and abroad, and its classification efficiency and accuracy is low. Based on the characteristics parameters of structure, shape and texture, the vehicle grille recognition method of the improved C-Support Vector Classification (C-SVC) based on the optimal parameters searching algorithm is proposed in this paper, where the efficiency and the precision are controlled by the dual-angle constraint: on the one hand, based on the Mahalanobis distance and a-principle, and combining with the weighted judgment, the sample data is sorted and used to accelerate the training and testing speed of the Support Vector Machine (SVM) and to improve the algorithm generalization efficiency; on the other hand, in the process of setting kernel function parameter, the optimal parameter iterative searching algorithm based on priori knowledge is designed to improve the classification accuracy of the classifier. The experiment shows that the accuracy rate of vehicle grille recognition method is 97.53%, representing the advantages of higher accuracy and lower false detection rate. It is also proved that this method is able to optimize the classification efficiency and to meet the real-time requirements of recognition.