基于兴趣点特征匹配的印刷图像缺陷检测

Defect Detecting of Printing Matter Based on Interest Points Feature Matching

提出了1种基于兴趣点特征匹配的印刷图像缺陷检测方法。首先利用SIFT特征检测器,获取图像中具有旋转和平移不变性的特征点,然后采用最近邻方法进行特征点匹配,最后使用最小二乘优化算法,计算2幅图像特征点对的变换矩阵来实现缺陷图像的检测。实验结果表明:该方法能够快速准确地提取出2幅图像间的对应特征点,大大降低了误匹配的概率,在图像发生较大平移和旋转的情况下,仍能准确地实现图像的对准及缺陷图像的检测。

A defect detecting of printing matter based on interest points feature matching was proposed. Firstly SIFT feature detector was used to gain the interest points which were invariant to image translation and rotation, then Nearest Neighbor was applied to extract the feature points pair, finally transforming matrix was found through least squares method to realize defect detecting. The experimental results indicated that this method can fast and accurately extract the corresponding feature points between two images and greatly reduce the probability of mismatching. Under the existing of significant translation and rotation situation, this method still can effectively realize the image registration and defect detecting.