基于局部区域SIFT特征匹配的偷盗检测算法研究

he scale-invariant feature extraction and SVD matching algorithm was studied, on this basic, a theft detection algorithm based on the limited area SVD-SIFT feature matching is proposed, for the detection of theft defection on the intelligent video surveillance system at the present time. What's more, we propose a iterative density-constraint algorithm for the purpose of eliminate the mismatch points maximally in the Feature matching results, to achieve that eliminate the mismatch points while reserving the correct matching point, at the same time, the positioning of detected objects fixed accurately. As the experimental results show, this method can monitor multiple objects and make sure them security at the real-time, it's more adaptable to complex environment, more effective, so the false and negative alarm rate is lower.