Eureka AIR delivers breakthrough ideas for toughest innovation challenges, trusted by R&D personnel around the world.

Defect detection method and device, model construction method and computer equipment

A technology of defect detection and defect location, which is applied in the field of image processing, can solve the problems of complex model parameters, difficult definition of features, inability to realize effective detection of surface defects of items or products, and achieve the effect of reducing the cost of use

Active Publication Date: 2021-04-02
锋睿领创(珠海)科技有限公司
View PDF3 Cites 4 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

However, due to the wide variety of defect types, the characteristics are difficult to define, and only occur in the production process, it is difficult to detect surface defects based on computer vision.
Moreover, most popular convolutional neural network models have complex network structures and a large number of model parameters in order to improve detection accuracy, which makes it difficult to deploy convolutional neural models on embedded systems with limited hardware resources and computing resources, and cannot realize Effective detection of surface defects on articles or products

Method used

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
View more

Image

Smart Image Click on the blue labels to locate them in the text.
Viewing Examples
Smart Image
  • Defect detection method and device, model construction method and computer equipment
  • Defect detection method and device, model construction method and computer equipment
  • Defect detection method and device, model construction method and computer equipment

Examples

Experimental program
Comparison scheme
Effect test

Embodiment 1

[0058]In this example, seefigure 1 It is shown that a defect detection method includes the following steps:

[0059]S100: Get the standard feature of the image to be detected.

[0060]The image acquisition system can be used to scan the item or product surface to collect items or product surface images, and the image is preprocessed to obtain the image to be detected. The pretreatment method includes cutting and scaling the collected item or product surface image, obtains an image of a predetermined size size, rotating, flipping, adjusting the image to a standard contrast, adjusting the brightness to a standard brightness. It will be appreciated that items or products include medical device products, industrial products, prints, product packaging and food packaging, etc., this embodiment is not limited.

[0061]The defect detection model to be detected to be detected is input to the pre-training standard, and the standard convolution to the detected image is standardized to obtain a standard...

Embodiment 2

[0092]In this example, seefigure 2 The step S300 shown by the defect detection method includes the following steps:

[0093]S310: Feature fusion of the Nth feature enhancement diagram and the M-characterizing method for feature fusion using the feature fusion layer of the defect detection model to acquire fusion feature, m <n.

[0094]Demonstrative, pre-constructed defect detection models can include six adaptive feature enhancements, as shown in the following table.

[0095]

[0096]Since the objective or product surface defect target is often small, the grayscale is not obvious, so the feature information of the defective target will disappear when the network reaches a certain depth, and therefore, according to the surface defects contained in each layer of the network The feature information determines the depth of the network, built an optimal network frame, as shown in the table below.

[0097]After the various layers of the neural network can be subjected to a reactionary layer, the converg...

Embodiment 3

[0109]In this example, seeFigure 4 A defect detection model constructor is shown, including:

[0110]S10: The adaptive feature enhancement layer is constructed by using the divided convolution layer, the multi-dimensional activation layer, and the point-volume layer.

[0111]A conventional convolution operation can be broken down into 3 × 3 due to fractal decomposition layer and 1 × 1 point volume layer, and a multi-portable activation layer is added between the divided convolution layer and the point-volume layer. , The feature fusion is made by each of the feature figures in the factor decomposition convolution layer to perform feature fusion, enhance the effective feature, and suppress the characteristics of the interference.

[0112]S20: The defect detection model is constructed using a plurality of adaptive feature enhancements.

[0113]S30: Enter the training image to the defect detection model.

[0114]S40: The visualization feature of the training image corresponding to each adaptive featu...

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More

PUM

No PUM Login to View More

Abstract

The embodiment of the invention discloses a defect detection method and device, a model construction method and computer equipment. The method comprises the steps of obtaining a standard feature map of a to-be-detected image; carrying out adaptive feature enhancement on the i1th feature enhancement graph by utilizing an ith adaptive feature enhancement layer of a pre-trained standard defect detection model so as to obtain an ith feature enhancement graph, and when i is equal to 1, taking the 0th feature enhancement graph as the standard feature graph and i is greater than or equal to 1 and less than or equal to N; and detecting the real defect position of the to-be-detected image according to the Nth feature enhancement image. According to the invention, defect detection on the surface ofthe product can be realized on an embedded system with limited hardware resources and computing resources by utilizing a new lightweight defect detection model, so that the use cost of the hardware resources and the computing resources is effectively reduced.

Description

Technical field[0001]The present invention relates to the field of image processing, and more particularly to a defect detection method, a device, a model constructor, and a computer device.Background technique[0002]All kinds of items or products are prone to various defects in the production process, which may affect the service life and reliability of the product, so surface defect detection is a key link of quality control. The surface defect detection method based on machine vision has the advantages of high efficiency, high accuracy, high real-time real-time, extensive research and application in the field of defect detection. However, due to the large number of defects, the characteristics are difficult to define, and only in the production process, it is difficult to encounter computer vision or product surface defect detection methods. Further, in order to improve the detection accuracy, most of the complicated network structures and a large number of model parameters, which...

Claims

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More

Application Information

Patent Timeline
no application Login to View More
Patent Type & Authority Applications(China)
IPC IPC(8): G06T7/00G06K9/62
CPCG06T7/0004G06T2207/20081G06T2207/20084G06F18/253G06F18/24G06F18/214
Inventor 何良雨崔健刘彤
Owner 锋睿领创(珠海)科技有限公司
Who we serve
  • R&D Engineer
  • R&D Manager
  • IP Professional
Why Eureka
  • Industry Leading Data Capabilities
  • Powerful AI technology
  • Patent DNA Extraction
Social media
Eureka Blog
Learn More
PatSnap group products