Looking for breakthrough ideas for innovation challenges? Try Patsnap Eureka!

Steel ladle number detection and recognition method based on Yolov3 algorithm

A recognition method and ladle technology, applied in character recognition, neural learning methods, character and pattern recognition, etc., can solve problems such as low accuracy, large image noise, and unintuitive

Pending Publication Date: 2021-05-18
HUNAN CHAIRMAN IOT INFORMATION TECH CO LTD
View PDF1 Cites 0 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

[0002] During the circulation of steel ladles in the steelmaking workshop, in order to correctly match the steel ladle with the molten steel information it carries and monitor the use status of the steel ladle, it is necessary to track and manage the steel ladle; in order to ensure the safe, orderly and efficient operation of the steel ladle, use the automatic Recognition technology is particularly critical; due to the harsh environment of the steelmaking workshop and the influence of high-temperature molten steel, the lighting conditions of the collected images are complex, and the dust in the air makes the image noise larger
[0003] The Chinese invention patent with the notification number CN110321751A discloses a method for identifying the steel ladle number, which is to carry out binary coding on the steel ladle number, then weld the code on the surface of the steel ladle, and finally collect the ladle image for scanning and identification; this method is also to collect the ladle image , can not eliminate the influence of factors such as illumination and noise in the image; and the method of using encoding has disadvantages such as unintuitive and difficult to maintain
[0004] Early image recognition technology was mainly based on the method of manually selecting features, combined with a shallow neural network classifier to achieve the effect of recognition; this type of method is very dependent on the prior knowledge of experts, and this method has a limited number of image features selected , cannot fully adapt to different environmental conditions, and the recognition accuracy is low

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
  • Steel ladle number detection and recognition method based on Yolov3 algorithm
  • Steel ladle number detection and recognition method based on Yolov3 algorithm
  • Steel ladle number detection and recognition method based on Yolov3 algorithm

Examples

Experimental program
Comparison scheme
Effect test

Embodiment 1

[0034] In this case, the ladle number template with the printed number is welded on the outer wall of the ladle; figure 2 As shown, the ladle number template is a circular iron template, and the corresponding number is deducted from the middle. After the template is welded in a suitable position, use a white high-temperature-resistant crayon to fill in the number gaps in the template, and draw on the outer circle. Circle; This can form a large color contrast with the background color of the ladle, and the welded template can ensure the long-term validity of the numbers; the purpose of drawing a circle outside is to only recognize the numbers in the circle and prevent the interference of other numbers.

[0035] When receiving the ladle arrival signal sent by the host computer, call the camera to take pictures and save the ladle image as training data; the arrival signal of the host computer is mainly judged according to the position and weight information of the crane ladle.

Embodiment 2

[0037] In this embodiment, after collecting more than 3,000 images, the Gaussian smoothing filter is first used to denoise the images, and the image denoising effect is as follows image 3 ;Gaussian smoothing filter makes the whole image uniform and smooth, removes details, and filters out noise; different from mean filter, the number in the kernel of Gaussian smoothing filter presents a Gaussian distribution, and the two-dimensional Gaussian distribution formula is as follows:

[0038] G(x,y)=1 / (2πσ^2)e^(-(x^2+y^2) / (2σ^2))

[0039] Among them, σ is a Gaussian parameter, which determines the width of the Gaussian filter; the larger the value of σ, the flatter the mode distribution map and the larger the template.

[0040] Then use the labeling tool labelImg to label the denoised image set. The information of each label includes "x", "y", "w", "h" of the corresponding pixel box, x, y represent the image The coordinates of a pixel point, that is, the coordinates of the upper le...

Embodiment 3

[0042] The Yolov3 algorithm uses the darkNet53 convolutional neural network as the basic network for feature extraction. The darkNet53 structure diagram is as follows Figure 5 , the Yolov3 algorithm is described as: (1) first divide the image into S*S fixed grids, if the center of the target falls on the corresponding grid, the grid is responsible for the detection of this target; (2) where each grid is Classify the B objects falling in the grid to obtain the probability of the category to which the object belongs, then regress to predict its location information, and finally integrate the location coordinates and category probability into a string of information; (3) Since each object contains multiple predictions Frame, use the non-maximum value suppression algorithm to filter out the frame with the highest probability of each object; (4) The output of the final network is the corresponding category and position information of each grid.

[0043] The Yolov3 loss function is...

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 steel ladle number detection and recognition method based on a Yolov3 algorithm, and the method comprises the following steps: 1, printing striking standard numbers on the outer wall of a steel ladle, and deploying a high-definition camera at a position where the steel ladle must pass through, so as to recognize the numbers printed on the outer wall of the steel ladle; 2, removing noise in the steel ladle number image by using a Gaussian smoothing filter, and marking the steel ladle number and the number in the image; 3, training a Yolov3 model by using the marked data set, and obtaining a group of network weights after the model is converged; step 4, using the trained network weight to calculate the ladle number and the number of the image to be identified. According to the invention, the steel ladle number does not need to be designed and coded, and visual numbers can be accurately recognized; the invention can well adapt to the complex environment of a steelmaking workshop, and deeper features of the image are extracted; the detection accuracy is high, and false detection and missing detection are few.

Description

technical field [0001] The invention relates to the technical field of iron and steel production process information technology, in particular to a method for detecting and identifying ladle numbers based on the Yolov3 algorithm. Background technique [0002] During the circulation of steel ladles in the steelmaking workshop, in order to correctly match the steel ladle with the molten steel information it carries and monitor the use status of the steel ladle, it is necessary to track and manage the steel ladle; in order to ensure the safe, orderly and efficient operation of the steel ladle, use the automatic Recognition technology is particularly critical; due to the harsh environment of the steelmaking workshop and the influence of high-temperature molten steel, the lighting conditions of the collected images are complex, and the dust in the air makes the image noise larger. [0003] The Chinese invention patent with the notification number CN110321751A discloses a method f...

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): G06K9/32G06K9/40G06K9/62G06N3/04G06N3/08
CPCG06N3/08G06V20/62G06V10/30G06V30/10G06N3/047G06F18/23213G06F18/241G06F18/2415
Inventor 陈勇波王利鑫钱艳萍李清源
Owner HUNAN CHAIRMAN IOT INFORMATION TECH CO LTD
Who we serve
  • R&D Engineer
  • R&D Manager
  • IP Professional
Why Patsnap Eureka
  • Industry Leading Data Capabilities
  • Powerful AI technology
  • Patent DNA Extraction
Social media
Patsnap Eureka Blog
Learn More
PatSnap group products

Browse by: Latest US Patents, China's latest patents, Technical Efficacy Thesaurus, Application Domain, Technology Topic, Popular Technical Reports.

© 2024 PatSnap. All rights reserved.Legal|Privacy policy|Modern Slavery Act Transparency Statement|Sitemap|About US| Contact US: help@patsnap.com