基于目标检测和语义分割的人行道违规停车检测  

Detection of Illegal Sidewalk Parking Based on Object Detection and Semantic Segmentation

作　　者：赵逸如 刘正熙[1] 熊运余[1] 严广宇 ZHAO Yi-ru;LIU Zheng-xi;XIONG Yun-yu;YAN Guang-yu(College of Computer Science,Sichuan University,Chengdu 610065)

机构地区：[1]四川大学计算机学院,成都610065

出　　处：《现代计算机》2020年第9期82-88,共7页Modern Computer

摘　　要：针对人行道车辆违停,提出一种人行道停车检测算法,使用基于深度学习的目标检测算法检测车辆目标,使用语义分割算法划分区域,通过判断目标位置是否处于人行道区域来判断是否为违规停车。使用混合高斯背景模型排除运动中车辆,提取干净背景,从而实现人行道停车实时检测。Aiming at the illegal parking of sidewalk vehicles,a sidewalk parking detection algorithm is proposed,which uses object detection algo rithm based on deep learning to detect vehicle targets,uses semantic segmentation algorithm to divide the area,and determines whether the target location is in the sidewalk area to determine whether it is illegal parking.The Gaussian Mixed Model is used to exclude moving vehi cles and extract a clean background to realize real-time detection of sidewalk parking.

关 键 词：违章停车 深度学习 目标检测 语义分割 混合高斯模型 

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