摘要
针对车型识别中计算时间过长、识别精度低的问题,提出基于整体与局部特征融合的车型识别方法。利用Sobel算子、Gabor滤波器组分别提取车辆的整体特征以及局部特征,结合灰关联分析对维数较高的局部特征进行优化;为减少稀疏表示中字典维度过高带来的计算耗时问题,提出双层串行分类策略,第一层中以整体特征为依据对车型进行粗略划分,第二层中以局部特征为依据实现最终的车型辨识。实验结果表明,该方法能有效提高车型识别的正确率,减少高维特征带来的计算时间消耗。
A vehicle type recognition based on global and local feature fusion was proposed aiming at solving the problem of long time consumption and low recognition accuracy in vehicle recognition.Sobel operator and Gabor filter were used to extract global feature and local feature respectively.Grey relational analysis was adapted to optimize the local feature.A two-layer serial classification strategy was proposed to reduce time consumption of sparse representation based on high dimension dictionary.In the first layer,the vehicle type was distinguished roughly according to global feature.Further vehicle type was recognized accurately by using local feature in the second layer.Experimental results show that the proposed method can improve the accuracy of vehicle type recognition and reduce the time consumption resulted from high dimension features.
出处
《计算机工程与设计》
北大核心
2016年第4期1051-1055,共5页
Computer Engineering and Design
基金
国家自然科学基金项目(61304205
61203273
41301037)
江苏省自然科学基金项目(BK20141002)
江苏省高校自然科学基金项目(13KJB120007)
大学生实践创新训练基金项目(201510300228
201510300276)
