Tobacco box character sticker extraction technology based on machine vision
Download as PDF
Zhenxun Jin, Yongye Jin, Wei Wang, Yaping Tan, Yong Zhang
With the introduction of the smart factory, the tobacco industry is facing an upgrade in the way of reading and verifying the package number in the process of material entry and exit. This paper proposes a method based on mechine vision to extract and recognize the characters on the surface of the tobacco box during the in and out of the warehouse. The proposed method uses image processing algorithms to perform a series of processing on the tobacco box image. Firstly, the target box and the background area are completely separated. Then the character printing type on the surface of the box is recognized, using vertex fitting and correction algorithm for complex distorted images, as well as overlapping sticker segmentation algorithms. The proposed method can efficiently locate and extract various different types of character regions from the tobacco package image to provide precise character lines for subsequent character recognition.
Machine vision; Character extraction; Distortion correction; Overlapping split