摘要
随着计算机及视频处理技术的快速发展,基于出租车的智能化非现场执法成为可能。但目前尚缺少针对出租车违法违章行为的智能分析技术,对此提出一种吸烟行为的自动检测算法。首先利用提出的亮度筛选规则减少图像增强部分的处理耗时;其次结合Haar-Adaboost和提出的分段直方图匹配算法,实现对出租车车窗区域的识别;设计一组有代表性的特征来识别吸烟烟雾和抖烟动作,包括烟雾质心运动轨迹、面积增长率、烟雾凸包与轮廓周长比、轮廓内外接矩形面积比以及抖烟频次和时间间隔,最后利用支持向量机进行特征分类。实验结果表明,上述方法的最优查准率达到85.7%,证明了算法的有效性。
With the rapid development of computer and video processing technology,intelligent off-site law en-forcement based on taxis has become possible,but there is still no intelligent analysis technology for taxi violations.In this regard,an automatic detection algorithm for smoking behavior is proposed.Firstly,the proposed brightness filtering rule was used to reduce the processing time of the image enhancement part.Secondly,the combination of Haar-Adaboost and the proposed piecewise histogram matching algorithm was used to realize the recognition of the taxi window area.A representative set of features was designed to identify the smoke and the smoking movement,in-cluding smoke centroid trajectory,area growth rate,smoke convex hull and contour circumference ratio,contour radii within the contour and the frequency and time interval of shake cigarette.Finally,support vector machine was used for feature classification.The experimental results show that the optimal precision of the method reaches 91.7%,which proves the effectiveness of the algorithm.
作者
黄训平
贾克斌
刘鹏宇
HUANG Xun-ping;JIA Ke-bin;LIU Peng-yu(Faculty of Information Technology,Beijing University of Technology,Beijing 100124,China;Beijing Laboratory of Advanced Information Networks,Beijing 100124,China;Beijing Key Laboratory of Computational Intelligence and Intelligent System,Beijing 100124,China)
出处
《计算机仿真》
北大核心
2020年第12期337-344,共8页
Computer Simulation
基金
国家自然科学基金面上项目(61672064)
北京市自然科学基金重点项目暨教委重点科技项目(KZ201610005007)
北京市自然科学基金面上项目(4172001)
北京市交通行业科技项目(2017058)
先进信息网络北京实验室(PXM2019_014204_500029)。
关键词
出租车
吸烟行为
自动检测
车窗区域
特征分类
Taxi
Smoking behavior
Automatic detection
Window area
Feature classification
