改进NCC算法在铝塑药板包装缺陷检测中的应用

Application of Improved NCC Matching Algorithm in the Detection of Packaging Defects of Aluminum Plastic Blisters

目的鉴于传统归一化互相关算法(Normalized cross correlation,NCC)存在计算量大及无法适应图像旋转的缺陷,提出一种改进算法,能快速完成对泡罩区域的提取,以及对铝塑药板缺损、漏装等缺陷的识别。方法首先对原始图像进行预处理,然后通过仿射变换使其变换至指定位置,接着提取其单个泡罩区域作为模板并构建积分图,最后获取待测图像,变换至相同位置后进行查表式匹配。结果与传统方法相比,改进后算法速度得到极大提升,对为1920×1200图片的匹配时间仅为21 ms,对实验样品的检测误检率为0,漏检率为3.5%.结论改进后的NCC匹配算法在满足精度要求的同时具有较快的速度优势,能较好地适用于铝塑泡罩包装缺陷检测中对泡罩区域的快速提取,药粒缺损10%以上缺陷及漏装缺陷的识别。

The work aims to propose an improved algorithm to quickly complete the extraction of blister area,and complete the defect identification of aluminum plastic plate,such as defect,missing,etc.in view of the shortcomings of traditional NCC algorithm,such as large calculation and failure to adapt to image rotation.Firstly,the original image was preprocessed and then transformed to the designated position by affine transformation.Secondly,a single blister area was extracted as a template and an integral map was constructed.Finally,the image to be measured was obtained,and then transformed to the same position for table matching.Compared with the traditional method,the speed of the improved algorithm was greatly improved.The matching time of 1920×1200 images was only 21 ms,the detection error rate of the experimental samples was 0,and the missed detection rate was 3.5%.The improved NCC matching algorithm not only meets the accuracy requirements,but also has the advantage of faster speed,which can be well applied to the rapid extraction of blister areas in the defect detection of aluminum plastic blister packaging,and the identification of defects with more than 10%of drug particle and missing loading defects.