结合目标检测和语义分割的违规广告牌检测  

Illegal Billboard Detection Based on Object Detection and Semantic Segmentation

作　　者：刘罗成 刘正熙[1] 熊运余[1] LIU Luocheng;LIU Zhengxi;XIONG Yunyu(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》2021年第12期127-132,共6页Modern Computer

摘　　要：针对城市中道路上的广告牌乱摆乱放的现象,提出一种基于深度学习的违规广告牌的检测算法。首先使用目标检测算法获取广告牌的位置,然后使用语义分割算法划分道路的区域,最后结合这两个算法的结果判断出广告牌是否处于人行道中间或者车行道,进而来判断是否属于违规摆放。In view of the phenomenon that billboards are placed in disorder on the roads of cities,a detection algorithm of illegal billboards based ondeep learning is proposed,which uses objection detection algorithm to detect the location of the billboard,uses semantic segmentation algo-rithm to divide road areas,and uses the results of the two algorithms to determine whether the billboard is in the middle of the sidewalk orin the roadway,and then to determine whether the billboard is placed illegally.

关 键 词：违规广告牌 深度学习 目标检测 语义分割 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]