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

Aviation wire rod recognition code automatic identification method based on convolutional neural network

A convolutional neural network, automatic recognition technology, applied in the field of image recognition, to achieve the effect of improving safety factor, reducing direct contact, and fast recognition speed

Inactive Publication Date: 2021-07-30
CHENGDU AIRCRAFT INDUSTRY GROUP
View PDF11 Cites 0 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

However, the training set is all images with clear and standard framing, which is only suitable for simple natural scenes such as shelves and small deformed text.

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
  • Aviation wire rod recognition code automatic identification method based on convolutional neural network
  • Aviation wire rod recognition code automatic identification method based on convolutional neural network
  • Aviation wire rod recognition code automatic identification method based on convolutional neural network

Examples

Experimental program
Comparison scheme
Effect test

Embodiment 1

[0025] This embodiment discloses a method for automatic recognition of aviation wire identification codes based on convolutional neural networks. As a basic implementation of the present invention, it includes data collection, data cleaning, data enhancement, network model training and decoding integration;

[0026] Data collection, use high-definition industrial cameras to collect image data of aviation wire identification codes, and transmit the collected image data to the host computer through the router, and use the host computer to perform image data to be recognized and the text information of the wire code contained in the image Collect to form a training set;

[0027] Data cleaning, using the upper computer to carry out sample training on the acquired image data. During the sample training process, there are label errors in the training set, vertically arranged text, excessive horizontal compression of the image, no identification code text in the image, and wires. Six...

Embodiment 2

[0033] This embodiment discloses a method for automatic recognition of aviation wire identification codes based on convolutional neural networks, as a basic implementation of the present invention, that is, in embodiment 1, in the process of data cleaning, for images of vertically arranged text , get the number n of characters in the image, the width l of the characters, and the overall width L of the characters arranged s , calculate the minimum width threshold L min =l*n, and judge L s Is it greater than L min ; if L s ≤ L min , then it is judged to arrange the text vertically; if L s > L min , it is judged as a non-vertical text; further, for the text occlusion caused by the wire, the number and continuous amount of the occluded text are judged, and the continuous text occlusion amount N c t Wire pictures < 4 are retained, and other occluded wire pictures are judged to be too large in occlusion, which is not suitable for automatic recognition; if the text is too blurr...

Embodiment 3

[0035] This embodiment discloses a method for automatic recognition of aviation wire identification codes based on convolutional neural networks, as a basic implementation of the present invention, that is, in embodiment 1, in the process of data enhancement, the enhancement process is performed by expanding the The image is randomly rotated from 1° to 180° and the brightness of the image is randomly changed to restore the disordered position of the wires that may exist in the real scene, identify the situation where the light is too bright or too dark, and incorporate adversarial samples to improve the safety factor of the algorithm .

[0036] This technical solution integrates adversarial samples into the training set for enhanced training, improves the safety factor of the model, and defends against external malicious attacks.

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 belongs to the technical field of image recognition, and particularly relates to an aviation wire rod recognition code automatic recognition method based on a convolutional neural network, which comprises the steps of collecting image data and forming a training set; performing sample training on the acquired image data, and cleaning six types of low-quality image data in a training set; performing enhancement processing on the cleaned image data; connecting a Bi-GRU circulation layer network in series behind a convolutional layer of a convolutional neural network CNN, integrating a Dropout layer at the tail end, and adopting a training strategy from easy to difficult to improve the recognition precision and reduce the training time; and finally, generating a wire code noun dictionary based on the wire code character sequence in the data set, and matching the recognition result with the noun dictionary to obtain a final recognition result. The technical scheme is based on the convolutional neural network, the recognition accuracy is improved, the model has high robustness and high recognition speed, and compared with a common end-to-end training scheme, the convolutional neural network is used as a main component of the network, and the degree of parallelism is high.

Description

technical field [0001] The invention belongs to the technical field of image recognition, and in particular relates to an automatic recognition method for aviation wire identification codes based on a convolutional neural network. Background technique [0002] The aviation wire identification code contains important information such as the manufacturing standard, type, specification, manufacturing country, manufacturer, and manufacturing date of the cable wire. The identification of the identification code is the basis for the use and maintenance of the wire. Under the current technical background, it is more common to use image recognition text technology. It is mainly divided into optical character recognition (Optical Character Recognition, OCR) and text recognition in natural scenes (Scene Text Recognition, STR). Optical character recognition OCR is mainly aimed at the optical character recognition of scanned documents, and is very mature in theory and application. Tex...

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): G06K9/20G06K9/62G06F40/216G06F40/289G06N3/04G06N3/08
CPCG06F40/289G06F40/216G06N3/08G06V10/22G06N3/045G06F18/214
Inventor 许艾明黎小华刘倍铭方亿李仁宏
Owner CHENGDU AIRCRAFT INDUSTRY GROUP