基于直线检测与密度聚类的车位识别算法研究  

Research on Parking Space Recognition Algorithm Based on Line Detection and Density Clustering

作　　者：曾俊琪 张足生 张声健 李文杰 ZENG Junqi;ZHANG Zusheng;ZHANG Shengjian;LI Wenjie(School of Cyberspace Security,Dongguan University of Technology,Dongguan 523808,China)

机构地区：[1]东莞理工学院网络空间安全学院,广东东莞523808

出　　处：《东莞理工学院学报》2023年第3期41-48,共8页Journal of Dongguan University of Technology

基　　金：国家自然科学基金面上项目(61872083);广东省普通高校重点领域专项(2020ZDZX3054)。

摘　　要：车位识别技术是实现智能停车管理的基础。现有的车位识别算法的识别图像多为环视图像,而针对高位图像的车位识别算法则较少,而且算法的召回率和精确率低。针对该研究现状,本文提出一种基于直线检测与密度聚类的车位识别算法,该算法能够准确识别不同高位图像中的车位。算法通过对检测直线进行密度聚类,得到了准确的车位线信息。同时,算法能够自动提取图像中的感兴趣区域,并准确定位感兴趣区域内的车位顶点,根据车位顶点坐标识别车位。实验结果表明,相比传统的车位识别算法,该算法具有更高的召回率和精确率。Parking space recognition technology is the basis of intelligent parking management.Most of the recognition images of the existing parking space recognition algorithms are panoramic images,while the parking space recognition algorithms for high posi⁃tion images are less,and the recall and accuracy of the algorithms are low.Aiming at the current research situation,this paper pro⁃poses a parking space recognition algorithm based on line detection and density clustering,which can accurately identify parking spaces in different high position images.The algorithm obtains accurate parking line information by processing the detected lines with density clustering.At the same time,the algorithm can automatically extract the region of interest in the graph and accurately locate the parking space vertexes in the region of interest,which are used to identify the parking spaces.The experimental results show that this algorithm has higher recall rate and precision rate than the traditional parking space recognition algorithm.

关 键 词：车位识别 直线检测 密度聚类 感兴趣区域 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]