基于改进几何活动轮廓模型和Kalman滤波的目标跟踪方法  

Motion object tracking algorithm using an improved geometric active contour model and Kalman filtering

作　　者：李晗[1] 武奇生[2] 罗向龙[1] 

机构地区：[1]长安大学信息工程学院,陕西西安710064 [2]长安大学电子与控制工程学院,陕西西安710064

出　　处：《长安大学学报（自然科学版）》2011年第3期90-94,共5页Journal of Chang’an University(Natural Science Edition)

基　　金：陕西省交通科技项目(08-13K);陕西省自然科学基金项目(SJ08-ZT13-8)

摘　　要：针对几何活动轮廓模型在跟踪时初始化的影响和收敛不准确的问题,将帧间差分和统计量假设后得到的目标外接矩形作为曲线初始值;引入一个强制项,提出一种改进的几何活动模型方法进行目标轮廓拟合,完成检测;结合目标物轮廓曲线和Kalman滤波器实现运动跟踪。试验结果表明:以车辆目标外接矩形作为初始化曲线,可简化初始化工作,加快车辆目标的轮廓曲线收敛速度;在收敛过程中引入了一个水平集函数的强制项,可使曲线准确演化到对象边缘的凹陷部分,增强曲线的收敛能力;在运动视频对象的准确轮廓基础上,可更准确地跟踪车辆目标。In order to overcome the initialization and inaccuracy problem of geometric active contour model in tracking, an improved method is proposed. The statistical feature of inter-frame difference is used to detect automatically the moving area of vehicles in a video sequence. Then this rectangle region of objects is set as the initialization of geometric active contour to fit the edge of objects, which can simplify the initialization. Lastly, an accurate contour of vehicles can be used for Kalman filtering tracking. Taking bounding rectangles as initial curve will simplify the initialization and improve the convergence rate. Through introducing an enforcement item to the level set function, the fitted contour can converge to concave edges, which also increase the convergence performance. The accurate contours can improve the tracking efficiency. 1 tab, 4 figs, 11 refs.

关 键 词：模式识别 运动目标跟踪 几何活动轮廓 水平集方法 KALMAN滤波 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]