中文摘要：

首先采用自适应帧差法对视频中运动目标进行检测,依据帧差面积的变化趋势选择采用2帧差法还是3帧差法来自动调节帧差面积,以防止检测目标的丢失;然后采用数学形态学方法对帧差图像进行后处理,根据帧差区域的质心移动推测运动目标的质心位置,同时将帧差区域拟合为与其具有相同标准2阶中心矩的椭圆,从而确定测地活动轮廓（GAC）模型的初始轮廓.水平集函数在归一化传导率非线性热方程的约束下,消除了耗时的重新初始化过程,减少了迭代次数和运行时间.实验结果表明了所提出算法的有效性.

英文摘要：

The adaptive frame difference method is used to detect moving objects in video sequences,in which two-or three-frame difference method is chosen to adjust the frame difference area automatically according to the changing trend of the frame difference area in order to prevent the loss of detected objects.Then the centroid position of moving objects is inferred according to centroid movement of frame difference region after post-processing using mathematical morphology on the frame difference image.A fitting ellipse which has same second-order central moments with the frame difference region is generated as the initial contour for the geodesic active contour（GAC） model.The level set function restricted by a nonlinear heat equation with normalized diffusion rate eliminates the time-consuming re-initialization procedure completely and reduces the number of iterations and the runnning time.Experimental results verify the effectiveness of the proposed algorithm.