Recognition of Tobacco Yellowing Degree in Curing Process Based on Deep Learning

In order to identify the Browning degree of tobacco leaves in the curing process, image acquisition equipment was installed in the curing houses of 11 tobacco producing areas in China, and the image data of tobacco leaves in curing process were obtained. The improved lightweight model EfficientNet-GECA is used for training and testing, the ordinary convolution in the MBConv module is replaced by the Ghost module, and the ECA attention mechanism is used to replace the SE attention mechanism, which reduces the amount of parameters and calculation of the model and improves the running speed. The recognition accuracy of the improved model in the test set reaches 88.74%, which is 1.61 percentage points higher than that before improvement. Compared with VGG16, ResNet50 and other models, the accuracy is increased by 9.76
^%
and 5.27
^%
respectively. The number of model parameters is 2.97M, which is only 56.14%and 11.62%of EfficientNet and ResNet-50, and the recognition speed of a single image is only 16.86ms. The improved EfficientNet-GECA model can recognize the Browning degree of tobacco in the baking process, which provides important support for the development of intelligent tobacco baking.