摘要
对区域生态及海表时空监测系统中的红外视频图像进行运动目标检测,提出鲁棒主成分分析(RPCA)和光流法结合的检测算法。对图像RPCA算法提取出的稀疏前景寻找特征点,利用金字塔Lucas-Kanade(LK)流法计算特征点并进行目标运动估计,得到目标运动的区域。再通过形态学分割得到最终的前景目标并进行跟踪。该算法在检测过程中避免背景像素点所带来的影响,消除背景减除法在运动目标提取过程中容易出现的"空洞"现象,弥补单独使用光流法检测耗时、计算复杂的缺陷。仿真结果表明,该算法具有鲁棒性优点,可应用于实际场景中,可以在具有复杂背景的环境中准确地提取出运动目标。
For moving target detection in infrared video images which is the spatial and temporal detection system of regional ecology and sea surface,proposes the optical flow method fused with RPCA（Robust Principal Component Analysis, RPCA）. The RPCA algorithm is used to extract the sparse foreground and find the feature points. In order to obtain the moving target area, uses pyramid Lucas-Kanade（LK） optical flow method to calculate feature points and estimate target movement. The final foreground goal is obtained and tracked through morphological segmentation. The algorithm avoids the influence of the background pixels in the process of moving object detection and eliminates the hollow phenomenon and makes up for the defection of time-consuming and complex calculation using optical flow algorithm separately. The simulation results show that the method proposed can detect and extract the moving target in the dynamic background with better robustness, real-time and integrity.
作者
于雯越
安博文
赵明
YU Wen-yue;AN Bo-wen;ZHAO Ming(College of Information Engineering,Shanghai Maritime University,Shanghai 201306)
关键词
鲁棒主成分分析(RPCA)
红外图像
运动目标
LK光流法
角点检测
Robust Principal Component Analysis （RPCA）
Infrared Image
Moving Target Detection
LK Optical Flow
Corner Detection
