作　　者： (张培）; (郭慧）; (刘芳辉）; (胡方尚）;

机构地区： 华东理工大学机械与动力工程学院,上海200237

出　　处： 《东华大学学报（自然科学版）》 2017年第4期592-596,602,共6页

摘　　要： 针对传统印刷品缺陷检测方法速度慢、检出率低的问题,提出一种仿射变换的快速检测方法.首先利用SURF(speeded-up robust feature)算法进行特征点匹配,并采用RANSAC(random sample consensus)算法剔除误配点后得到高纯度的匹配点对,然后利用最小二乘法拟合得到变换参数,最后通过图像差分和后处理,识别并标记出缺陷所在位置.经过试验验证,相对模板匹配和分层检测算法,该方法无论在算法速度还是缺陷检出率上都有很大提高. Aiming at the problem of the slowness and low detection rate of the defect detection methods of traditional printed matter, a fast detection method of affine transformation is proposed. Firstly, the feature points are matched by SURF（speeded-up robust feature） algorithm. RANSAC（random sample consensus） algorithm removes the mismatched points and obtains the matching point pairs with high purity. Then, the transformation parameters are obtained by least squares fitting. Finally, the position of defects is identified and marked by image difference and postprocessing. Experiments show tha~： the proposed method is much better than the template matching algorithm and the hierarchical detection algorithm in terms of algorithm speed and defect detection rate.

关 键 词： 算法 算法 仿射变换 缺陷检测 印刷品