← 返回
基于迁移学习的深度卷积神经网络的孟加拉国车辆分类与检测
Bangladeshi Vehicle Classification and Detection Using Deep Convolutional Neural Networks With Transfer Learning
| 作者 | Manish Kumar Dwivedi · R. Jayapragash |
| 期刊 | IEEE Access |
| 出版日期 | 2025年1月 |
| 技术分类 | 储能系统技术 |
| 技术标签 | 储能系统 PFC整流 |
| 相关度评分 | ★★★★★ 5.0 / 5.0 |
| 关键词 | AC - DC转换器 功率因数校正 总谐波失真 电池充电 电路结构 |
语言:
中文摘要
车辆分类检测是深度学习和图像处理在智能交通管理和AI辅助驾驶中的重要应用。本文提出孟加拉国车辆分类检测系统,实现低速高速车辆检测。测试了YOLOv8、MobileNetV2等11种预训练CNN模型在六个数据集上的性能,发现YOLOv8 Classify、MobileNetV2和GoogLeNet表现最佳。改进LabelImg标注工具并采集达卡市5460张图像的54556个标注对象,涵盖16类车辆。部署YOLOv8 Detect和SSD-MobileNet V2到NVIDIA Jetson Nano,实现93%检测率和98%准确度。
English Abstract
This paper introduces an AC-DC semi-bridgeless dual-switch SEPIC converter, specifically designed for battery charging applications. It focuses on improving power factor (PF) by substantially reducing total harmonic distortion (THD) in the AC input current through a modified converter structure. Additionally, the converter achieves significant reductions in the sizes of its inductors as it operates in discontinuous conduction mode (DCM) to achieve the low current THD. This converter topology employs two power switches to realize the power factor correction (PFC). Primary novelty of the proposed converter lies in designing and selecting a circuit structure which ensures low THD and unity power factor by the energy balance principle of inductors and capacitors. Additionally, the incorporation blocking diodes in the proposed converter effectively eliminates circulating current through input inductor which in turn makes the converter more efficient. This novel circuit structure also eliminates the requirement of additional closed loop control algorithm for PFC. To validate the proposed concept, a prototype converter of 100W/53V is developed and tested. This converter yields a current THD of 2.1%, unity PF and efficiency of 92.4% at rated condition during hardware testing. The paper also conducts a comparative analysis with similar other converters to assess the performance of the proposed solution thoroughly. This comprehensive study underscores the effectiveness of the new converter for PFC and reducing input current THD for battery charging applications.
S
SunView 深度解读
该车辆识别技术可应用于阳光电源新能源汽车充电场景。阳光在充电桩和充电站建设中,车辆识别算法可优化充电资源分配,实现车牌识别、车型识别和智能调度。结合阳光OBC和充电桩产品,该深度学习技术可提升充电站运营效率,支持V2G车网互动,实现电动汽车有序充电和削峰填谷功能。