基于ViT的高速公路车辆细分类研究

Research on Freeway Vehicle Fine-Grained Classification Based on ViT

对外观相似但类别不同的车辆进行识别,一直都是车辆细分类问题的主要研究内容。为了有效地识别相似车辆之间具有细微差距的特征,提出了基于ViT的车辆细分类模型,模型中的车辆部件选择模块可以较好地选择车辆图片中重要的特征,从而排除图片中无关区域对分类的影响。实验表明,与传统的分类方法相比,该方法在高速公路车辆细分类问题上取得不错的效果。

Recognition of vehicles for similar appearance but different categories has always been the main research content of vehicle fine-grained classification.In order to effectively to identify the features of slight gaps between similar vehicles,a vehicle fine-grained classification model based on ViT is proposed.The vehicle component selection module in the model can better select the important features of the vehicle image and exclude the influence of irrelevant areas in the image of the classification.Experi⁃ments show that compared with traditional classification methods,this method achieves good results in fine-grained classification of freeway vehicles.