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

Method for identifying casting DR image loose defects based on improved YOLOv3 network model

A network model and defect technology, applied in the field of workpiece casting, can solve the problems of affecting the local area effect of defects, it is not easy to distinguish the types of objects to be detected, and the recognition accuracy is not high

Pending Publication Date: 2020-07-31
CHONGQING UNIV
View PDF0 Cites 12 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

[0003] Problems in the above methods: Method 1) basically processes the image globally, which affects the effect of the local area of ​​the defect, and it is not easy to distinguish the type of object to be detected. Due to the large number of features in the image, defect recognition is also vulnerable noise disturbance
In method 2), the traditional machine learning network model (Bayesian classifier, support vector machine) or shallow neural network model is currently used to realize DR image defect detection based on DR image defect detection of castings with complex features. The recognition accuracy is not high compared to the deep learning framework

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
  • Method for identifying casting DR image loose defects based on improved YOLOv3 network model
  • Method for identifying casting DR image loose defects based on improved YOLOv3 network model
  • Method for identifying casting DR image loose defects based on improved YOLOv3 network model

Examples

Experimental program
Comparison scheme
Effect test

Embodiment 1

[0071] see Figure 1 to Figure 6 , a method based on the improved YOLOv3 network model to identify loose defects in casting DR images, mainly including the following steps:

[0072] 1) Obtain DR loose defect images of several castings.

[0073] The casting is a cast steel bolster or a side frame of a railway train bogie.

[0074] 2) Preprocessing the DR loose defect image and constructing a loose defect data set.

[0075] The main steps of preprocessing the DR original defect image are as follows:

[0076] 2.1) The DR original defect image is uniformly divided into defect images of N×N size.

[0077] 2.2) Perform data enhancement on defect images. The data augmentation methods include image flipping, image rotation and mirroring.

[0078] 3) Preprocessing the loose defect data set, enhancing the gray value of the loose defect data set, making it easier to distinguish from the gray value of the background, thereby improving the process of subsequent training models and tes...

Embodiment 2

[0140] A method for identifying loose defects in casting DR images based on the improved YOLOv3 network model, mainly comprising the following steps:

[0141] 1) Obtain DR loose defect images of several castings.

[0142] 2) Preprocessing the DR loose defect image and constructing a loose defect data set.

[0143] 3) Preprocessing the loose defect data set to enhance the gray value of the loose defect data set.

[0144] 4) Use the rectangular frame of the image annotation tool to mark the defect of the loose defect data set, and obtain the defect category, defect level, center point coordinates (X, Y) of the rectangular frame, rectangular frame width W and rectangular frame corresponding to each rectangular frame Height H. The sparse defect data set after defect labeling is randomly divided into a training set of sparse defect data and a test set of sparse defect data.

[0145] 5) Establish an improved YOLOv3 network model, and set the number of filters, COCO data set and V...

Embodiment 3

[0152] A method for identifying loose defects in casting DR images based on the improved YOLOv3 network model, the main steps are shown in Example 2, wherein the main steps of training the improved YOLOv3 network model are as follows:

[0153] 1) Divide each image of the loose defect data training set into s×s cells;

[0154] 2) Use the improved YOLOv3 network model to extract features from each cell, and produce feature images of 3 different scales;

[0155] 3) Using the regressor to predict several candidate target bounding boxes, the main steps are as follows:

[0156] 3.1) Set preset bounding box (c x , c y ,p w ,p h ); said (c x , c y ) is the center coordinate of the preset bounding box on the feature image; p w ,p h is the width and height of the preset bounding box on the feature map;

[0157] 3.2) Calculate the center offset of the predicted bounding box (t x , t y ) and aspect ratio (t w , t h );

[0158] 3.3) Update the predicted target bounding box (b...

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 method for identifying casting DR image loose defects based on an improved YOLOv3 network model. The method comprises the following steps: 1) performing defect labeling on aloose defect data set by utilizing a rectangular frame of an image labeling tool; and 2) establishing an improved YOLOv3 network model; 3) training the improved YOLOv3 network model by using the loosedefect data training set; 4) testing the trained improved YOLOv3 network model by using the loose defect data test set; 5) improving the improved YOLOv3 network model, and 6) obtaining a DR image ofthe casting to be detected, inputting the DR image into the improved YOLOv3 network model, and judging the defect grade and position coordinates of the casting. According to the invention, detection effects of small target objects by a target detectio network are improved.

Description

technical field [0001] The invention relates to the field of workpiece casting, in particular to a method for identifying loose defects in DR images of castings based on an improved YOLOv3 network model. Background technique [0002] The purpose of casting DR image defect recognition is to find out the position of the defect from the casting radiographic image, then extract various information of the defect, and finally complete the identification of the defect. At present, there are three main methods for casting defect identification: 1) direct detection based on image processing; 2) traditional machine learning model detection based on defect location and tracking; 3) casting DR based on deep learning frameworks such as Faster Rcnn Detection of image defects. [0003] Problems in the above methods: Method 1) basically processes the image globally, which affects the effect of the local area of ​​the defect, and it is not easy to distinguish the type of object to be detect...

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
IPC IPC(8): G06T7/00G06T7/11G06T7/62
CPCG06T7/0008G06T7/11G06T7/62G06T2207/10116G06T2207/20081G06T2207/20084G06T2207/30116G06T2207/30164
Inventor 段黎明阮浪杨珂朱世涛
Owner CHONGQING UNIV
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