Design And Implementation Of Label Detection System Based On Machine Vision

With the development of society, labels are playing an increasingly important role in people’s work and life and people are also paying more attention to the quality of labels. Because of the influence of production technology and mechanical precision, many quality problems often occur when the labels are producing, such as, the label damaged, poor printing including the more print, the less print, short stroke and scratches. Then label defects detection is a vital link in our lives.The topic is developing a set of label defects detection system for an air conditioning manufacturer. This system can detect the labels’ quality automatically for the manufacturer. The whole process of defect detection can be handled automatically by the software without human’s effort. The operators only need to confirm the defect identified. It can combine the manual work and the intelligence so as to improve the detection rate of the labels and assure the labels’ quality. The primary work of this article includes:(1) According to the system design requirements and pain points, it make a general design for the whole labels defect detection system including the system program, the technological line, the selection of hardware equipment and software division of functional modules.(2) Design the label image preprocessing algorithm, including image noise reduction, standard template production and image registration process. By analyzing and researching, the paper eventually adopts fast median filtering algorithm for noise reduction, standard template based on image registration algorithm and image registration algorithm based on template matching.(3) Design the label image defect detection and classification algorithms. This paper introduces the defect detection algorithm based on image difference. It can detect the defects on the labels well. Then the paper adopts defect tag algorithm based on equivalence classes of travel so as to tag each defect target area. Finally, the paper uses Euclidean distance classifier to identify each defect type.(4) The paper eventually finishes the label defect detection system with the help of VC++6.0 for software programming development tools. We test the system in the lab, the results have showed that the system’s accuracy, precision and real-time performance can meet the user’s requirements.