基于RPCA低秩模型的车辆外形识别研究  

Vehicle Shape Recognition Based on RPCA Low-Rank Model

作　　者：樊锐博 景明利 李岚 魏玉峰 FAN Rui-bo;JING Ming-li;LI Lan;WEI Yu-feng(School of Electronics Engineering,Xi'an Shiyou University,Xi'an Shaanxi 710065,China;School of Science,Xi'an Shiyou University,Xi'an Shaanxi 710065,China)

机构地区：[1]西安石油大学电子工程学院,陕西西安710065 [2]西安石油大学理学院,陕西西安710065

出　　处：《计算机仿真》2023年第6期132-137,244,共7页Computer Simulation

基　　金：国家自然科学基金(61673314);陕西省重点研发计划(2020GY-152);西安石油大学博士创新基金(290088266);西安石油大学研究生创新与实践能力培养计划资助(YCS21112066)。

摘　　要：车辆外形识别作为车辆分类的依据具有重要现实应用价值。目前基于神经网络的车辆识别方法严重依赖于训练样本数量与质量,同时受限于训练时间长、泛化能力弱和可解释性差等不足。针对车辆外形识别问题,基于RPCA低秩模型,提出了利用低秩模型的车辆外形识别算法。首先,上述算法利用RPCA低秩模型提取出视频中目标物体的稀疏特征信息;然后,运用方向梯度直方图提取车辆的特征矩阵;最后,采用SVM分类器实现车辆外形的分类。实验结果表明,在较少的训练样本下,所提算法识别率或者准确率优于经典的PCA+SVM和Harr+AdaBoost算法,是一种可行的车辆外形识别算法。As the basis of vehicle classification,vehicle shape recognition has important practical application value.Current vehicle recognition methods based on neural networks rely heavily on the number and quality of training samples,and are limited by the long training time,weak generalization ability,and poor interpretability.Ai-ming at the problem of vehicle shape recognition,based on the RPCA low-rank model,a vehicle shape recognition al-gorithm using the low-rank model is proposed.First,the algorithm uses the RPCA low-rank model to extract the sparse feature information of the target object in the video;then,uses the directional gradient histogram to extract the feature matrix of the vehicle;finally,the SVM classifier is used to classify the vehicle shape.The experimental results show that with fewer training samples,the recognition rate or accuracy of the proposed algorithm is better than the classical PCA+SVM and Harr+AdaBoost algorithms,and it is a feasible vehicle shape recognition algorithm.

关 键 词：鲁棒主成分分析 方向梯度直方图 支撑向量机 车辆识别 

分 类 号：TP391.9[自动化与计算机技术—计算机应用技术]