摘要
针对道路视频监控中局部特征车辆品牌和型号识别率低的问题,提出了一种离散粒子群优化的识别算法。用形态学定位法提取视频中车前脸区域,能够快速获得识别的感兴趣区域。提取车前脸的SURF特征作为识别局部特征,对视角变化和光线变化有较好的鲁棒性。在离散环境下定义粒子的位置和速度,设计粒子的更新规则,利用离散粒子群优化获得待识别图像特征点在标准图像中的最佳覆盖,提高特征点匹配的正确率,从而提高车型识别的正确率。最后利用具有对应关系的特征的相似度进行对比识别。建立了15种车系76种车型的车前脸图像库进行实验,实验结果表明改进方法的车型正确识别率为93.6%。
In order to improve the recognition rate of vehicle brand and model, this paper presented a vehicle-type recognition method based on discrete particle swarm optimization algorithm in road video surveillance. To begin with, a vehicle frontal face was determined by the method of mathematical morphology,which was identified more quickly. Then SURF features were extracted from the vehicle front face that has better robustness to changes of perspective and light. Using discrete particle swarm optimization, the best coverage of the extracted features in the standard image was obtained by the new definition of position and velocity of particles in the discrete environment and rules of updating particles for the sake of improving the accuracy of feature points matching and vehicle recognition. Finally the charac- teristic similarity of the corresponding features was used for recognition. Experiments were carried out based on a dataset of vehicle frontal face images including fifteen vehicle makes and seventy-six models. The experimental results show that the correct recognition rate of the proposed method is 93.6%.
出处
《计算机仿真》
CSCD
北大核心
2016年第1期177-180,共4页
Computer Simulation
基金
南京航空航天大学研究生创新基地(实验室)开放基金(kfjj201428)
国家自然科学基金(61371170)
关键词
车辆品牌和型号
离散粒子群优化
拓扑结构
车前脸
Vehicle brand and model
Discrete particle swarm optimization
Topology structure
Vehicle frontal face
