基于仿射变换与模板匹配的车位识别与计数算法研究  

Research on parking spaces recognition and counting algorithm based on affine transformation and template matching

作　　者：应捷[1] 陈文[1] 杨海马[1] 何施晶 刘瑾[3] Ying Jie;Chen Wen;Yang Haima;He Shijing;Liu Jin(School of Optical Electrical&Computer Engineering,University of Shanghai for Science&Technology,Shanghai 200093,China;Shanghai Lingjing Information Technology Co.,Ltd.,Shanghai 201306,China;School of Electronic&Electrical Engineering,Shanghai University of Engineering Science,Shanghai 201620,China)

机构地区：[1]上海理工大学光电信息与计算机工程学院,上海200093 [2]上海领旌信息科技有限公司,上海201306 [3]上海工程技术大学电子电气工程学院,上海201620

出　　处：《计算机应用研究》2022年第3期919-924,共6页Application Research of Computers

基　　金：国家自然科学基金联合基金资助项目(U1831133)。

摘　　要：针对道路停车位识别与计数的问题,对基于仿射变换与模板匹配的方法进行了研究。首先对采集的视频图像进行预处理,并采用基于PPHT与直线聚类的方法检测停车位候选区并计算出车位标线的倾斜角;然后采用基于仿射变换与模板匹配的算法检测停车位入口处标线并用红色圆标记;接着采用感知哈希算法剔除相似度较高的图像,将剩余的图像按顺序连接起来;最后采用动态规范法、Lagrange函数和Grubbs准则记录连接图中停车位数目。实验中分别在晴天、阴天和雨天采集四个不同路段的停车图像制作了相应的数据集,并对所提出的方法进行了定量评估。结果表明,该方法车位识别的准确率为98.46%,计数的准确率为99.52%。Aiming at the problem of road parking space recognition and counting, this paper studied the method based on affine transformation and template matching.Firstly, it preprocessed the collected video images, and detected the candidate par-king spaces based on PPHT and line clustering, and calculated the inclination angle of parking lot marking.Then, it used an algorithm based on affine transformation and template matching to detect the marking at the entrance of the parking space and mark it with a red circle.Then, it used the perceptual hash algorithm to eliminate the images with high similarity and connected the remaining images in order.Finally, it used the dynamic gauge method, Lagrange function and Grubbs criterion to record the number of parking spaces in the connection diagram.In the experiment, it collected the parking space images of four different sections in sunny, cloudy and rainy days to make the corresponding data sets, and the proposed method was evaluated quantitatively.The results show that the accuracy of the proposed method is 98.46% and the accuracy of counting is 99.52%.

关 键 词：车位识别 直线聚类 仿射变换 模板匹配 感知哈希 

分 类 号：TP393.04[自动化与计算机技术—计算机应用技术]