摘要： 针对传统违章停车人工检测方式准确率低、成本高等缺点，提出一种基于视频监控的检测算法。利用码本模型对背景进行建模，用以检测监控场景中的运动目标，通过计算联通区域面积来滤除其他物体对监控车辆的干扰，并根据车辆停留时间将车辆分类，一旦发现违停车辆，系统会发出报警。实验结果表明，该算法准确率高、实时性好。

关键词: 视频监控, 违章停车, 码本模型, 车辆分类, 噪声滤除, 实时性

Abstract: Aiming at the shortcomings of traditional illegal parking detection methods such as low accuracy rate and high costs, this paper proposes a detection algorithm based on video surveillance. The moving objects are detected in video by using a codebook background modeling algorithm. The interference of other objects is filtered by computing the area of the connected region. Vehicles are classified according to the parking time. If illegal parked vehicles are found, the system will alarm. Experimental results demonstrate that the algorithm obtains high accuracy and achieves real-time performance.

Key words: video surveillance, illegal parking, codebook model, vehicle classification, noise filtering, real-time performance

中图分类号: 

• TP391