Abstract: A new approach is proposed to detect and classify the moving vehicle in static scenes,which is based on GVF-Snake model and inertia ellipse.The vehicles contour is extracted from successive traffic-frames by Gaussian Mixture Model,Expectation Maximization estimate algorithm and improved GVF-Snake model.The ellipse eccentricity of the moving vehicle is computed from the principal of inertia ellipse of rigid body.The vehicle classifier is established on the base of three parameters,which are the length of vehicle,the projection area of the vehicle and the equivalent ellipse eccentricity.As the method is based on the principles of statistics,it is quite simple and it can be easily programmed.This method is fit to be applied into the embed system.

Key words: GMM (Gaussian Mixture Model), GVF-Snake (Gradient Vector Flow-Snake) Model, eccentricity

摘要： 提出了一种在静止背景交通图像序列中运动车辆的检测和分类方法,即基于GVF-Snake模型和惯量椭圆的车辆分类算法。利用混和高斯模型(GMM)、期望最大化(EM)估计算法、改进GVF-Snake模型,从序列交通视频图像中检测出运动车辆;然后,借用刚体惯量椭圆原理,计算运动车辆等效椭圆偏心率,从而建立车长-车投影面积-车的等效椭圆偏心率三参数建立了车辆分类器。该方法的车辆检测与分类都是基于数理统计原理,算法复杂度小,可用数字逻辑编程实现,适合在嵌入式系统中应用。

关键词: 混和高斯模型(GMM), GVF-Snake模型, 偏心率