基于交错帧差和高斯模型的运动车辆检测

Moving vehicle detection based on interframe difference and Gaussian model

针对现有运动车辆检测方法中存在的不足,提出一种基于交错帧差和改进的高斯模型的运动车辆检测方法。通过检测道路边界提取感兴趣区域并使用均值法建立初始背景模型,对得到的感兴趣区域进行自适应分块,进行交错帧差实现背景的粗分割,在此基础上利用改进的高斯模型对背景进行精分割,运用双阈值背景差分对精分割结果进行后处理,得到运动车辆。实验验证该方法能够对运动车辆实现快速、准确的检测。

Aiming at the problems of the moving vehicle detection, an algorithm on the basis of an improved interframe differen-tial algorithm and an improved Gaussian model was presented According to a statistical histogram, an interesting region was ex-tracted. Using a mean algorithm, an initial background model was established The interesting region was divided into several blocks using a self-adaptive method. According to an improved interframe difference algorithm, the interesting region was sepa-rated roughly. On the basis of these steps, an improved Gaussian model was utilized to separate the rough results precisely. The results were processed by double-threshold background subtracting. Experimental results show this algorithm can detect moving vehicles rapidly and accurately.