摘要
为实现链条装配过程中链板表面商标的在线检测 ,提出了满足实时、通用性要求的商标识别方法。首先将部分软件功能硬件化 ,迅速获得图像的边缘信息 ,然后将图像边缘拆分成具有不同属性特征的小线段 ,进而利用加权统计的方法获得有效边缘线段的数量 ,并将其与评判阈值进行比较 ,以确定链板表面有无商标。以 0 8B链板为例进行了试验研究 。
In order to distinguish the different faces of link plates on line, an approach to automatic brand recognition of link plate was developed in this paper, which is applicable for various kinds of brand images. A hardware was used, instead of calculating software, to detect edges of brands rapidly, and each of the detected boundaries was divided up into several subsections with certain geometrical characteristics. With the help of weighted statistics, we got the sum of effective lines included in an image, and compared it with a given threshold value to ascertain whether there is a brand on link plate. The results of the experiments show that the method adopted in the paper is practical. With boundary division and weighted statistics, the amount of useful marginal information of an image is increased greatly, which makes it easy for computer to give a correct judgement. In practice, the recognition productivity is up to 4 pieces per second,which meets the requirement of a link assembling machine.
出处
《农业机械学报》
EI
CAS
CSCD
北大核心
2002年第4期91-94,共4页
Transactions of the Chinese Society for Agricultural Machinery
