基于深度学习的车辆计数算法的研究及优化  

Research and Optimization of Vehicle Counting Algorithm Based on Deep Learning

作　　者：张双贺 ZHANG Shuanghe(College of Computer Science and Technology,Southwest Jiaotong University,Chengdu 610000)

机构地区：[1]西南交通大学信息科学与技术学院,成都610000

出　　处：《现代计算机》2021年第22期106-111,共6页Modern Computer

摘　　要：传统的图像中车辆计数一般基于目标检测的方法,但是其在车辆较密集,遮挡较严重的场景下效果不佳。本文采用CrowdCounting中密度估计的方法,将车辆计数作为回归任务,学习图像内容和车辆密度分布之间映射。本文基于VGGNet提出AAMNet神经网络模型,结合注意力机制和金字塔结构,在车辆数据集High_Vehicles上取得了不错的效果。Traditional vehicle counting in images is generally based on the object detection method,but it does not work well in scenes with dense vehi-cles and severe occlusion.This paper adopts the density estimation method in Crowd Counting,and takes vehicle counting as a regression task to learn the mapping between image content and vehicle density distribution.This paper proposes a neural network model named AAM-Net based on VGGNet,which combines the attention mechanism and pyramid structure to achieve good results on the vehicle data set High_Vehicles.

关 键 词：深度学习 车辆计数 密度估计 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程] TP391.41[自动化与计算机技术—控制科学与工程] U495[交通运输工程—交通运输规划与管理]