Curdy and greasy coating classification algorithm based on convolutional neural network

Abstract

The invention is applicable to the technical field of respective tongue coating curdy and greasy identification, and provides a curdy and greasy coating classification algorithm based on a convolutional neural network; the convolutional neural network is used for performing feature extraction on curdy and greasy coating pictures, and can effectively combine the color, shape and texture of curdy and greasy coating to describe rich tongue coating image features; a possible curdy and greasy coating area is predicted through priori knowledge; the tongue picture is cut into a plurality of sub-images; therefore, the feature extraction method only needs to pay attention to the related sub-images instead of the whole image; because the multi-example classification decision is only decided by a minimum negative example in the normal tongue image and a maximum positive example in the curdy and greasy tongue image, and some overall information is obtained, the multi-example classification decision forces the positive packet (tongue image) to contain at least one curdy and greasy sub-image, and then whether the curdy and greasy coating is contained in the tongue image or not is judged by usinga multi-example learning method.

Description

Technical field [0001] The present invention relates to the technical field of separate identification of tongue fur and greasy fur, in particular to a classification algorithm for furnishing greasy fur based on convolutional neural network. Background technique [0002] Greasy tongue coating is one of the important tongue characteristics. At present, most modern researches on the recognition of tongue coating are based on computer intelligent information processing technology. [0003] At present, most methods of tongue image feature extraction are manual extraction methods. Manual extraction of features cannot effectively reflect the salient features of stale and non-sicky fur. Some scholars have also proposed to use convolutional neural networks to extract tongue image depth features. Located in the middle or root of the tongue, the rest of the information can be ignored, and this method focuses on the overall information of the tongue, thus capturing more useless information an...

AI Technical Summary

Problems solved by technology

[0004] The purpose of the embodiments of the present invention is to provide a convolutional neural network-based classification algorithm for putrid moss, which aims to solve the problem of low classification performance in the prior art

Images