Small-sample defect identification method based on deep learning

Abstract

The invention discloses a small-sample defect identification method based on deep learning. The method comprises: a step of performing small sample data enhancement, namely a step of performing data enhancement by using a digital image processing method and a generative adversarial network method; a step of inputting the enhanced data into a migration deep learning model feature extractor, and training a new classification structure to obtain a migration deep learning model set; a step of inputting testing data into the migration deep learning model, performing model fusion in a Voting model fusion mode, voting according to the principle that minority obeys majority, and thus a defect recognition result is obtained. According to the disclosed method, a deep learning model fusion thought and combination are utilized, the deep learning model DAGAN acts on data enhancement, and therefore, the deep learning model DAGAN can be applied to a deep learning recognition model only through a small number of samples, and higher accuracy can be achieved.

Description

technical field [0001] The invention relates to the technical field of industrial wood surface defect detection, in particular to a small-sample defect recognition method based on deep learning. Background technique [0002] In the wood industry, various types of wood are widely used in daily life. However, in the process of wood processing and production, defects such as cracks, worm holes, nodules, browning and rot on the surface of sawn timber are caused by differences in the growth environment, climate conditions, and transportation and processing processes. These defects will seriously affect the quality of wood products and cause waste of wood resources. Efficient defect identification technology is one of the important links to ensure the quality of wood processing products. [0003] At present, most domestic timber production still adopts artificial sampling method for defect detection, which is completed by subjective judgment by naked eye observation. This method...

AI Technical Summary

Problems solved by technology

[0004] In order to solve the problems existing in the prior art, the purpose of the present invention is to provide a small-sample defect recognition method based on deep learning. The present invention adopts the method of combining fusion thinking and deep learning to solve the problem that the data of wood surface defects is less and data collection is difficult. And labeling is time-consuming and labor-intensive, and it is difficult to use deep learning methods to identify wood surface defects

Images