Polarizer defect image real-time classification method based on parallel module deep learning

Abstract

A polarizer defect image real-time classification method based on parallel module deep learning belongs to the field of material defect detection, and comprises the following steps: 1, preparing a polarizer image data set; 2, building a deep learning network; 3, inputting the polarizer data set into a built deep learning network, extracting multi-scale features of the polarizer image through network training, and inputting the extracted features into a Softmax layer for classification to obtain a classification model; and 4, inputting the test image into the classification model, inputting theprobability that the image belongs to a certain category and the label corresponding to the image into an Accuration layer, and outputting a correct classification result of the image. According to the invention, image classification and a model compression method are combined by using deep learning; a polarizer defect image real-time classification network based on parallel module deep learningis built, the depth model is minimized and the detection speed is increased on the premise of not reducing the classification accuracy, and the real-time requirement of defect detection in the actualindustry is met under the condition of limited hardware resources.

Description

technical field [0001] The invention belongs to the technical field of material defect detection, and in particular relates to a method for real-time classification of polarizer defect images based on a parallel module deep learning network. Background technique [0002] The polarizer is one of the core components of the LCD panel, accounting for about 10% of the cost of the LCD panel. In the production process of polarizers, due to factors such as processing technology limitations, insufficient design level, production equipment failures and poor production conditions, uneven areas are easily formed inside the workpiece. These areas usually appear as bubble-like residual glue, cracks, Defects such as inclusions, stains, scratches, etc. Any tiny polarizer defects will be displayed on the display after being imaged by liquid crystal molecules. Human eyes are very sensitive to such local abnormalities in the display, which will affect the perception and reduce the quality of ...

AI Technical Summary

Problems solved by technology

But they all have their own shortcomings: Manual inspection is mainly through visual scanning of polarizers on the production line to classify defective products for subsequent processing

However, in the process of mass production, the detection accuracy and speed are easily affected by the subjective factors and experience of the inspectors, and it is difficult to meet the requirements of modern assembly lines; the traditional machine vision detection defects mainly process the image of the inspected object. In the process, it is necessary to manually define and select the feature representation that can accurately identify the defects in the image

But in an industrial setting, when a new problem arises, new features must be manually designed, due to the randomness, shape diversity, and complexity of defect areas and locations, the standard feature descriptors used to describe defects often lead to classification results Inaccurate, it is difficult to meet actual industrial requirements

Images