基于统计模型和GVF-Snake的彩色目标检测与跟踪

为了能使传统监视系统具备目标自动检测与跟踪能力,提出了一种基于统计模型和GVF（gradient vectorflow）-Snake的彩色目标检测与跟踪算法.该算法可用于解决在静态背景下通过彩色视频信息来对运动目标进行自动检测与跟踪的问题,同时可直接给出目标轮廓的数学表示,并可简化后续目标识别算法的设计.该算法首先采用归一化RGB空间与灰度空间相结合的模型取代单一灰度模型来消除阴影对目标检测的影响;接着在此模型的基础上对差分图像进行GMM（Gaussian mixture model）建模,并构造运动边界图像,然后将静态图像轮廓提取算法GVF-Snake引入运动图像中,并通过修改能量项,使其能够跟踪运动目标的轮廓,最后针对Snake初始轮廓需要手工设定的问题,提出一种根据目标区域自动初始化轮廓的方法,为加快GVF-Snake的收敛速度,还采用一阶差分算法来预测下一时刻目标轮廓的位置.实验结果证明,该算法对刚性和非刚性两类目标都有较好的跟踪效果,可应用于智能监视和交通监控等领域.

Statistical Model and GVF-Snake Based Color Targets Detection and Tracking

In order to enable traditional monitor systems to detect and track moving objects, a statistical model and GVF- Snake based algorithm is proposed, which employs color video information in static background to detect and track moving objects and provide a good representation of the objects so as to simplify the subsequent object recognition. The algorithm proposed replaces the plain gray space model with normalized RGB space combined with gray space model to eliminate the effect of shadow upon detection, A GMM of the difference of 2 successive frames is constructed, based on which, a motion border image is generated. GVF-Snake is then enhanced to extract the contours of moving objects in video sequence by modifying the energy entry and adding a method to initialize the Snake automatically. In order to accelerate the convergence of the Snake, the contour of next moment is predicted at every moment by estimating the center of the moving object using a 1st-order difference algorithm. This algorithm has been proved to be effective for both rigid and non-rigid objects and can be used for smart surveillance and traffic monitoring,