Embedded Inkjet Detection Based on a Convolutional Neural Network

Authors

  • Chenye Han School of Networks and Communications, Hebei Polytechnic Institute, Shijiazhuang 050091, China
  • Baojing Liu Information Technology and Cultural Management Institute, Hebei Institute of Communications, Shijiazhuang 051430, China

Abstract

In a high-speed production line, due to the large number of products and other factors, the inkjet printing of information about the product is easily lost or blurred. To ensure the integrity of inkjet information and eliminate non-acceptable products, an embedded inkjet detection model based on improved CNN is studied and constructed. YOLOv5s algorithm is used to locate the inkjet character area. In the course of research, it was found that the positioning algorithm still has problems such as low accuracy and slow reasoning speed. Therefore, an ECA attention mechanism and CIOU loss function are introduced to improve the algorithm. Then a convolution neural network combined with a cyclic neural network and CTC algorithm is used to recognize inkjet characters. It can timely detect missing information, ambiguity and other defects in the inkjet code. According to the
experiment, the average defect detection accuracy of the model is 97.56%, the Re value is 98.56%, and the running time is 26.25ms, which can achieve efficient and accurate inkjet-defect detection. The model provides a technical guarantee for industrial production quality inspection and has positive implications for the development of industrial production enterprises.

Keywords: convolution neural network; inkjet detection; character recognition; attention mechanism

Cite As

C. Han, B. Liu, "Embedded Inkjet Detection Based on a Convolutional Neural Network",
Engineering Intelligent Systems, vol. 31 no. 5, pp. 369-378, 2023.








Downloads

Published

2023-09-01

Issue

Vol. 31 No. 5 (2023): International Journal of Engineering Intelligent Systems

Section

Article

License

The submission of a paper implies that, if accepted for publication, it will not be published elsewhere in the same form, in any language, without the prior consent of the publisher. Before publication, authors are requested to assign copyright to CRL Publishing Ltd. This allows CRL to sanction photocopying, and to authorize the reprinting of issues or volumes according to demand. Authors' traditional rights will not be jeopardized by assigning Copyright in this way, as they retain the right to reuse the material following publication, and to veto third-party publication.