Unlock instant, AI-driven research and patent intelligence for your innovation.

A flaw detection method based on deep learning network

A technology of deep learning network and recognition method, which is applied in the field of recognition based on deep learning network to detect defects, can solve the problem that small objects cannot achieve an ideal effect, and achieve reduced computational complexity, improved accuracy, and high-precision defects Detection effect

Active Publication Date: 2022-07-12
ZHEJIANG GONGSHANG UNIVERSITY +1
View PDF1 Cites 0 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

However, the algorithms in the above two directions cannot achieve an ideal effect in the detection of small objects. On the basis of the deep learning one-stage model, the present invention improves the existing algorithm and proposes a method suitable for defective points. The detection method improves the detection rate of centimeter-level flaws and small targets, and at the same time increases the detection speed

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
  • A flaw detection method based on deep learning network
  • A flaw detection method based on deep learning network
  • A flaw detection method based on deep learning network

Examples

Experimental program
Comparison scheme
Effect test

Embodiment

[0030] A flaw identification method based on a deep learning network, comprising the following steps:

[0031] Step 1. Shoot a video image sequence containing defect points, and input it into the Resnet-50 network for feature extraction, specifically:

[0032] First, input the picture into the Resnet-50 feature extraction network, and then obtain a feature map of 7*7*2048, and then pass a convolution with a convolution kernel size of 1, a stride of 1, and a convolution number of 256 to reduce the feature. The number of channels in the graph, the feature graph after convolution is 7*7*256.

[0033] Step 2. Flatten the output feature map, add position encoding information, and put it into the transformer-encoder (encoder), specifically:

[0034] The flattening operation is as follows: change the shape of the feature map from 7*7*256 to 49*256, that is, change H*W*C to (H*W)*C, compress the height and width into the same dimension, and pass The flattened feature map, denoted as...

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 invention discloses a flaw point identification method based on a deep learning network. In the present invention, the defects are firstly extracted through the Resnet-50 feature extraction network to extract the characteristics of the defects, and then the defects are detected through the improved transformer network to identify the defects. By improving the transformer network module of the DETR network, the invention can improve the speed and detect defects more accurately.

Description

technical field [0001] The invention belongs to the field of image processing and target detection in computer vision, and relates to a recognition method for detecting defects based on a deep learning network. Background technique [0002] The traditional target detection is a technique of generating a proposal frame, then extracting the features in the target frame, and finally classifying it. Traditional object detection algorithms have bottlenecks in speed and accuracy. With the rapid development of deep learning neural network algorithms, object detection tasks in videos and images have also developed rapidly. [0003] Defect detection is a very important job in the industry. At present, it mainly relies on workers to manually select defects, which is very time-consuming and high labor costs. At the same time, workers are required to have rich experience. Existing object detection based on deep learning can be divided into the following two categories: one-stage and tw...

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 Patents(China)
IPC IPC(8): G06V20/40G06V10/82G06V10/764G06K9/62G06N3/04G06N3/08
CPCG06N3/084G06N3/044G06F18/2414G06F18/2415
Inventor 王慧燕姜欢
Owner ZHEJIANG GONGSHANG UNIVERSITY