基于奇异值分解的自适应模板更新研究

Adaptive template-updating through singular value decomposition

研究了基于奇异值分解的图像匹配和目标跟踪问题。由于图像的奇异值特征具有良好的稳定性,可以将奇异值当作一种有效的代数特征来描述并表征图像。根据所定义的奇异值缩放不变量提出了一种基于奇异值分解的模板更新算法。在算法中,根据奇异值向量的缩放不变特征来度量当前模板内的目标信息,然后根据所定义的置信度自动计算更新后所需的模板大小,从而使更新后的模板更有效地包含目标。试验表明:提出的模板更新算法在序列图像的目标跟踪中具有较好的实用性。

The problem of image matching and target tracking based on singular value decomposition （SVD） was discussed. The SVD had robust performance that was invariant to image disturbance and it made the singular value credible to represent the image as an algebraic feature. A template-updating strategy was proposed to update the current template based on the scale invariant character of the singular value vector. The information of the target in the current template was calculated firstly. Then the updated template that containing the accurate target was adaptively acquired according to the singular value＇s scale invariance and confidence degree. Experimental results show that the proposed strategy is practical and efficient in target tracking.