Abstract: An improved optical flow estimation algorithm based on the frame difference is proposed, which is used for detecting the moving target. In the first part, the discontinuous frame difference is applied to detecting the moving target with the inter-frame displacement smaller than one pixel and accumulated displacement bigger than one pixel. In the second part, the General Dynamic Image Model（GDIM） is introduced to form a new optical constraint equation, which has overcome the problem that the equation is not tenable. Only not zero pixels are calculated by using the optical flow method. As a result, accuracy and speed of the target detection are improved.The effectiveness of the proposed method is verified by the simulation results.

Key words: target tracking, frame difference, optical flow, local optical flow

摘要： 结合运动目标检测帧差法运算速度快和光流法活动目标检测准确度高的特点，提出一种改进的帧间差光流场计算的运动目标检测算法。在帧差部分采用隔帧差分从而可以检测到帧间位移小于1个像元而多帧累积位移大于1个像元的运动点目标；在光流计算时，引入通用动态图像模型（GDIM）建立新的光流约束条件，克服了亮度变化引起的约束方程不成立问题。算法仅对帧差法后图像中不为零的像素进行光流场计算，提高了目标检测的准确性和检测速度。仿真实验证明了该算法的有效性。

关键词: 目标跟踪, 帧差法, 光流法, 区域光流检测