基于机器视觉的收藏币类型识别及币面缺陷检测  

Collectible Coin Type Recognition and Coin Surface Defect Detection Based on Machine Vision

作　　者：罗彰 张从鹏 侯静娴 LUO Zhang;ZHANG Congpeng;HOU Jingxian(College of Mechanical and Material Engineering,North China University of Technology,Beijing 100144,China)

机构地区：[1]北方工业大学机械与材料工程学院,北京100144

出　　处：《机械工程师》2022年第1期12-14,共3页Mechanical Engineer

基　　金：北京市教委科技创新服务(2020YFF0304900)。

摘　　要：以开发收藏币币面品相检测系统为目标,研究了图像滤波、变换、类型识别及配准等技术。通过预处理初步分割出币面区域,采用多目标相关性模板匹配法,识别出收藏币类型及币面特征在图像中所处的位置,然后根据匹配结果将目标图像与参考图像进行配准,在配准的基础上进行差分,实现对收藏币币面缺陷区域的提取。再根据不同类型缺陷的灰度、几何、形状特征的不同对缺陷区域进行分类识别,然后统计缺陷区域的面积、坐标等参数。实验数据表明,收藏币币面图像识别配准高效准确,能够有效提取出缺陷区域并准确统计缺陷参数。With the goal of developing a collection coin face detection system, this paper studies the techniques of image filtering, transformation, type recognition and registration. Preliminary segmentation of the currency area through image preprocessing, the type of the coin and the position of the coin feature in the image are confirmed by multi-target relevance template matching method. Then the target image and the reference image are registered according to the matching results. The extraction of the defect area of the coin surface of the collectible coin is realized by the difference based on the registration. The defect area are classified and identified based on the region gray, geometric, and shape characteristics of different types of defects. Then the area, coordinates and other parameters of the defect area are counted and saved. Some experiments are carried out. The image recognition and registration of the coin face of the collection coin is efficient and accurate, which can effectively extract the defect area and accurately count the defect parameters.

关 键 词：图像处理 图像差分 图像分割 缺陷提取 

分 类 号：TP274[自动化与计算机技术—检测技术与自动化装置]