Abstract: As traditional intelligent traffic detection systems have poor real-time capability and can be easily influenced by illumination variations, which leads to a problem in identifying vehicles peccancy and capturing the image about them. A robust and fast vehicles peccancy detection and tracking method is proposed based on color difference histogram algorithm and Kalman filter. Background fuzzy matching method is used to select initial background images. Color difference histogram algorithm which is robust to environmental change is used to detect moving targets. Kalman filter is used to track and predict the mass centroid of the moving targets in order to detect the same target in the predictive zone. The trajectory of the mass center is used to identify them and capture them. The experimental results through the detection of the actual roadway scene under different environments demonstrate that the proposed method can identify vehicles peccancy  fast and accurately.

Key words: intelligent traffic surveillance, color difference histogram, Kalman filter

摘要： 针对传统的智能交通系统中违章车辆检测方法实时性差、易受光照变化条件变化制约，影响后续辨别车辆违章和图像取证抓拍的问题，提出了一种基于颜色差分直方图和卡尔曼滤波的鲁棒、快速的违章车辆检测跟踪算法。该算法采用背景模糊匹配思想，选择初始背景图像；利用对环境变化鲁棒的颜色差分直方图算法检测运动目标；对运动目标团块的质心运动状态采用卡尔曼滤波进行跟踪预测，从而在预测的区域内检测同一目标团块；通过判断其质心运动轨迹，达到辨别违章车辆检测与抓拍的目的。通过对真实道路中不同天气条件下的场景进行检测，实验结果表明该算法能够快速而准确地检测违章车辆。

关键词: 智能交通监测, 颜色差分直方图, 卡尔曼滤波