Title: Research on digital online inspection of product quality in tobacco production enterprises

Authors: Feng Chen; Minye Liao; Chunning Deng; Ling Su

Addresses: Quality Management Department, Longyan Tobacco Industrial Co., Ltd., Longyan, 364021, China ' Quality Management Department, Longyan Tobacco Industrial Co., Ltd., Longyan, 364021, China ' Quality Management Department, Longyan Tobacco Industrial Co., Ltd., Longyan, 364021, China ' Quality Management Department, Longyan Tobacco Industrial Co., Ltd., Longyan, 364021, China

Abstract: Nowadays, many cigarette appearance quality problems in tobacco production enterprises lead to the frequent occurrence of unqualified cigarette products, and at present, there is a lack of automatic detection methods for cigarette appearance quality. In order to solve this problem, the deep convolution network model is applied to the automatic cigarette appearance quality detection system. First, the optimal parameters of the deep convolution network model are determined, and the optimal parameters of the model are: learning rate 0.01, maximum pooling function and relu activation function. The quality detection system is applied to the actual detection, and the results show that the accuracy rate of the automatic detection system of the deep convolution network model is 98.54%, and the false detection rate is 2.5%, which are better than the traditional manual sampling method. The above results show that the deep convolution network model does have high accuracy for automatic detection of cigarette appearance, and provides a new research idea for online detection of tobacco product appearance quality.

Keywords: tobacco appearance quality; CNN network; automated detection; VGG network; ReLU activation function; Sigmoid function; Gaussian filter.

DOI: 10.1504/IJISD.2024.140845

International Journal of Innovation and Sustainable Development, 2024 Vol.18 No.5/6, pp.629 - 647

Received: 13 Jul 2022
Accepted: 25 Oct 2022
Published online: 03 Sep 2024 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article