基于弱光环境的车辆识别研究  

Research on vehicle detection based on low light environments

作　　者：张峻祎 丁冰 丁洁[2] ZHANG Junyi;DING Bing;DING Jie(College of Electrical and Power Engineering,Taiyuan University of Technology,Taiyuan 030024,China;School of Integrated Circuits and Electronics,Beijing Institute of Technology,Beijing 100081,China)

机构地区：[1]太原理工大学电气与动力工程学院,山西太原030024 [2]北京理工大学集成电路与电子学院,北京100081

出　　处：《现代电子技术》2024年第7期17-24,共8页Modern Electronics Technique

基　　金：国家重点研发计划(2022YFB3204600);北京理工大学青年教师学术启动计划。

摘　　要：由于图像在弱光环境下具有低曝光、前景背景融合、对比度低等问题,因此难以在弱光环境下有效、实时检测图像中的目标车辆。目前为了提高检测效果,通常需要设计较为复杂的神经网络结构或建立额外的对比数据集,但这不仅降低了网络速度,也会提升网络训练成本。为了解决这一问题,提出一种弱光环境下的车辆识别网络,在该网络中一方面设计了特征提取模块与特征融合模块以提高网络在弱光环境下的检测能力,另一方面使用模块结构重参数化的方法以提高网络的检测速度。实验表明,该网络可在保证检测速度的同时有效地识别弱光环境下的车辆。Since low light environment images show the problems of low exposure,foreground background fusion,and low contrast,it is difficult to effectively detect the vehicles(the objects)in the image in real time.Usually,it is required to design more complex neural network structures or establish additional comparative datasets to improve the image detection effects.However,these methods not only reduce the speed of the network,but also increase the training cost of the network.Therefore,a vehicle recognition network in low light environments is proposed.On the one hand,a feature extraction module and a feature fusion module are designed to improve the network's detection ability in low light environments.On the other hand,a module structure reparameterization method is used to increase the detection speed of the network.Experimental results show that the network can effectively identify vehicles in low light environment images while ensuring detection speed.

关 键 词：弱光环境 车辆检测 神经网络 特征提取 特征融合 模块结构重参数化 

分 类 号：TN911.73-34[电子电信—通信与信息系统] TP391[电子电信—信息与通信工程]