84 / 2024-08-26 07:26:57

A classification method for surface cleaning of turnover bins based on an improved convolutional neural network

surface cleanliness detection; deep learning; improved convolutional neural network

Tobacco vacuum rewetting is a very important part in the process of tobacco production, after the tobacco is sent into the rewetting machine using a turnover box, it is necessary to go back to the production line to reload the turnover box, and it is necessary to keep the surface of the turnover box with a high degree of cleanliness during the reloading process. In order to address the problem of low recognition accuracy caused by too many network layers easily losing information during the recognition process of previous turnover box surface cleanliness detection using the convolutional neural network method, a multi-connected weighted alternative to the original CNN network is now proposed. The method mainly realizes fusion feature extraction by improving the CNN network, using multi-connected paths, and feature weighting of the original paths after multi-scale feature extraction of the turnover box images. The pictures of clean turnover box and dirt-containing turnover box are actually taken and collected on the experimental setup through multiple industrial cameras. By conducting experiments on this dataset, it is demonstrated that on this data, the use of the improved CNN method in extracting the dirt features of the turnover box can improve the surface dirt feature extraction capability of the ordinary CNN network, reduce the loss of information and improve the accuracy of the correct recognition of the cleanliness of the turnover box pictures.