基于小波提升的变模板视频多尺度目标识别*

为了快速有效地解决图像序列中移动目标的识别定位,提出了一种基于互相关的目标检测算法。该算法经小波提升建立低分辨率子图,通过快速的互相关检测,在高分辨率原图中建立紧凑ROI。为了适应运动目标变形,整个检测过程以互相关运算极大值为依据,更新目标模板。实验结果表明,在满足检测速度的同时,算法对目标的识别有较高的精度与鲁棒性。

Multi scale variable template target detection in image sequence based on lifting scheme

To detect a moving target in a video rapidly and accurately, proposed a target detection algorithm based on correlation. It constructed decomposition images by lifting scheme, and a correlation calculation can build a compact region of interest in the high resolution image rapidly. To apply this algorithm to deformable target detection, the target template would be update when the correlation level was below the threshold. Experimental results show that the algorithm has a good validity and robustness in target detection.