基于特征增强和分组模块的车型精细识别  

Fine Vehicle Recognition Based on Feature Enhancement and Grouping Module

作　　者：郑秋梅[1] 曹文龙 王风华[1] ZHENG Qiumei;CAO Wenong;WANG Fenghua(School of Computer Science and Technology,China University of Petroleum(East China),Qingdao 266580)

机构地区：[1]中国石油大学(华东)计算机科学与技术学院,青岛266580

出　　处：《计算机与数字工程》2024年第5期1406-1411,共6页Computer & Digital Engineering

基　　金：国家自然科学基金项目(编号:52074341,51874340)资助。

摘　　要：针对车型种类多、差异小,模型复杂,识别精度低的问题,提出一种基于特征增强和分组模块的车型精细识别方法,在ResNet网络基础上改进,在卷积块中加入多尺度通道域和空间域的注意力机制,增强对重要的特征提取,并将多通道特征图进行分组,根据分组损失函数不断优化分组,通过加权方式结合KL(Kullback-Leibler)散度损失函数和交叉熵损失函数,有助于网络学习类内差异小、类间差异大的特征。该方法在Stanford cars-196数据集和自制数据集上进行测试,验证了所提模型的有效性。In order to solve the problems of many types of vehicles,small differences,complex models and low recognition ac-curacy,a fine vehicle recognition method based on feature enhancement and grouping module is proposed,which is improved on the basis of ResNet network.The attention mechanism of multi-scale channel domain and spatial domain is added to the convolution block to enhance the extraction of important features,and the multi-channel feature graphs are grouped and continuously optimized according to the grouping loss function.KL(Kullback-Leibler)divergence loss function and cross entropy loss function are com-bined by weighted method.The method is tested on Stanford cars-196 dataset and self-made dataset to verify the effectiveness of the proposed model.

关 键 词：车型精细识别 多尺度 注意力机制 特征增强 损失函数 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]