Online grading method for tobacco based on deep learning algorithm

Abstract

The invention relates to an online grading method for tobacco based on a deep learning algorithm, which is characterized in that positive and negative classification, three-classification of positive group tobacco, green tobacco and sundry tobacco and then grading of the positive group tobacco are performed through a complete set of tobacco grading system, and the method is applicable to online grading; background masking is performed based on a method of supervised learning, and the background removing accuracy is higher compared with a traditional algorithm; classification and grading are performed by using all positive tobacco, and tobacco grading characteristics are more reflected on the positive tobacco, so that classification and grading performed by using the all positive tobacco have more rigorous scientificity and higher accuracy; grading for the positive group tobacco is performed by adopting a GoogLeNet model, and higher grading accuracy can be acquired by using more convolutions and larger number of network layers. The online grading method can perfectly achieves quick, lossless and online rejection of the green tobacco and the sundry tobacco, accurately divide the grade of the positive group tobacco and ensure all aspects of benefits of purchasing to be not damaged.

Description

Technical field [0001] The invention relates to an online tobacco grading method, in particular to a tobacco online grading method based on a deep learning algorithm. Background technique [0002] At present, many researchers use infrared spectroscopy or image processing technology combined with traditional machine vision algorithms, such as support vector machines, for tobacco leaf grading. There are few reports on online tobacco grading based on RGB images combined with deep learning methods. [0003] The concept of deep learning comes from the research of artificial neural networks. It was first proposed by Hinton et al. in 2006. Based on the Deep Belief Network (DBN), an unsupervised greedy layer-by-layer training algorithm is proposed, which brings hope to solve the optimization problems related to the deep structure, and then the deep structure of the multi-layer autoencoder is proposed. [0004] In addition, the convolutional neural network proposed by Lecun et al. is the fi...

AI Technical Summary

Problems solved by technology

[0002] At present, many researchers use infrared spectroscopy or image processing technology combined with traditional machine vision algorithms, such as support vector machines, to classify tobacco leaves. There are few reports on the online classification of tobacco leaves based on RGB images combined with deep learning methods.

Images