摘要
针对动态背景下的小目标检测问题,提出了基于双向稀疏光流融合的目标检测方法。首先采用FAST方法提取当前帧图像中的角点,然后在连续的三帧图像中进行前、后向稀疏光流跟踪,确定正确匹配的特征点对,利用匹配的特征点对计算用于背景补偿的帧间运动参数,最后在背景补偿的基础上进行三帧差分,以检测出图像中的运动小目标。实验结果显示,本算法能够很好地解决背景和目标同时快速运动的问题,为运动目标的跟踪奠定基础。
A target detection method to detect small target in dynamic background based on the fusion of bidirectional sparse optical flow is proposed. At first, corners in current image are extracted based on FAST method, and then bidirectional sparse optical flow between continuous frames are executed to determine matching points, and uses these pairs of matching points to calculate motion parameters between frames. Finally, after executing background compensation, frame difference is done to detect small moving target in image. The experimental resuhs show that this algorithm can well deal with rapid motion both of the background and targets.
出处
《电视技术》
北大核心
2016年第3期122-125,129,共5页
Video Engineering
关键词
FAST
动态背景
稀疏光流
帧差
背景补偿
FAST
dynamic background
sparse optical flow
frame difference
background