改进Haar-like特征联合CamShift算法用于车辆跟踪

Vehicle Tracing by Using Modified Haar-like Features and Cam Shift Algorithm

提出一种改进的Haar特征获取方法,并结合CamShift算法提取车辆检测信号实现特定车辆实时跟踪。首先利用倾斜45°特征与像素和的商计算信号的Haar特征值;其次利用AdaBoost算法训练样本得到特征信号分类器,构建特征信号样本级联分类器,利用获得的级联分类器对视频信号进行检测;最后将检测结果中的特定车辆外切矩形作为CamShift算法的初始窗口,并对CamShift算法后续跟踪窗口进行检测,以提高检测效率和跟踪实时性。实验表明,该方法对车辆图像信号具有较好的检测效果,对特定车辆跟踪具有较高的实时性。

For the problem of vehicle detection and real-time tracking of specific vehicle in intelligent traffic system,an improved Haar feature combined with CamShift algorithm is proposed for vehicle detection and real-time tracking of specific vehicle.Firstly,the Haar features are calculated by using the quotient of tilted 45°feature and the pixel sum,secondly,the AdaBoost algorithm is trained by the samples to obtain the classifier,and a feature sample cascade classifier is constructed to detect video images.Finally,specific vehicle tangent rectangle in the detection result is used as the initial window of the CamShift algorithm,and the CamShift algorithm subsequent tracking window is only detected to improve detection and tracking efficiency.Experiments show that the method has perfect detection effect and high real-time performance for specific vehicle tracking.