Automatic Detection of Welding Defects Using Faster R-CNN

基于快速R-CNN的焊接缺陷自动检测

简介：使用Inception-ResNet模型进行缺陷检测

数据集：射线图像

Abstract

Experts are required to properly detect the test results and it takes a lot of time and cost to manually Interpret the radio-graphic testing image of the structure over 500 blocks.

专家需要正确检测测试结果，手动解释超过500个区块的结构的无线电图形测试图像需要大量时间和成本。

a method of automatically detecting welding defect using Faster R-CNN which is a deep learning basis.using data augmentation method to artificially increase the limited data.

提出一种基于深度学习的快速R-CNN自动检测焊接缺陷的方法。采用数据扩充的方法，人为地增加有限的数据。

1. Introduction

For the weld testing, there are various technologies such as radiographic testing (RT), ultrasonic testing (UT) and magnetic testing (MT) used as non-destructive testing (NDT).

对于焊缝的检测，有射线检测(RT)、超声波检测(UT)和磁性检测(MT)等多种无损检测技术。

Since welding information of more than 2000 locations per block is manually prepared, omissions and errors commonly occur, which requires additional work, resulting in a huge amount of time and cost.

由于每个区块2000多个位置的焊接信息都是人工编制的，经常会出现遗漏和错误，这需要额外的工作，导致大量的时间和成本。

Of various deep learning algorithms, a convolutional neural network (CNN) that has recently been researched a lot for image classification shows high performance compared to conventional algorithms.

在各种深度学习算法中，卷积神经网络(CNN)在图像分类中表现出了比传统算法更高的性能。

not only the boundaries of defects but also the types of defects are important because NDT rules depending on the type of defect.that handles radiographic images, research on object detection employing CNN has already been underway.automatically detects the welding defects in radiographic images by employing Faster R-CNN that shows high-performance in terms of accuracy.

不仅缺陷的边界很重要，缺陷的类型也很重要，因为无损检测的规则取决于缺陷的类型。为了处理射线图像，利用CNN进行物体检测的研究已经在进行中。通过使用速度更快的R-CNN自动检测射线图像中的焊接缺陷，该R-CNN在准确性方面表现出高性能。

We compared ResNet [12] and Inception-ResNet V2 [13] that showed high-performance in ImageNet by configuring them as backbone networks.

我们通过将ResNet[12]和Inception-ResNet V2[13]配置为主干网络，比较了它们在ImageNet中表现出的高性能。

Table 1 describes the method and features of defect detection.

表1描述了缺陷检测的方法和特点。

2 Methodology

2.1. Convolutional Neural Network

2.2. Faster R-CNN

Faster R-CNN, as shown in Figure 2a, proposed a novel method by constructing a neural network in the conventional selective search as a method of obtaining region proposals.

Faster R-CNN，如图2a所示，提出了一种新的方法，通过在传统的选择性搜索中构建神经网络来获得区域建议。

2.3. ResNet

In forming a deep network, the gradient value becomes too large or saturated with small values, resulting in a vanishing gradient problem that loses or slows the learning effect.

在形成深度网络时，梯度值变得过大或被较小的值饱和，导致梯度消失问题，从而丢失或减慢学习效果。

ResNet added an identity shortcut connection to the conventional neural network structure to obtain the learning effect of the deep network.

ResNet 在传统的神经网络结构上增加了一个身份快捷连接，以获得深度网络的学习效果。

2.4. Inception-ResNet V2

Inception-ResNet is a model that combines structural features and is divided into V1 and V2.Figure 3b shows the module A of Inception-ResNet [13].

INSTIMATION-ResNet是一种结合了结构特征的模型，分为V1和V2。图3b显示了初始-ResNet的模块A[13]。

The module form of Inception-ResNet V1 and Inception-ResNet V2 are the same, but there are differences in the number of internal filters and the modification of stem.

Inception-ResNet V1和 V2的模块形式相同，但在内部过滤器的数量和stem的修改上有所不同。

The Inception-ResNet model improves performance due to the difference between Inception V3 and V4. The high recognition rate and learning rate are verified through recent studies, and it is expected to achieve high outcomes when used as the feature extractor of the welding defect detection algorithm.

由于Inception V3和V4之间的差异，Inception-ResNet模型提高了性能。近期的研究验证了该算法具有较高的识别率和学习率，将其作为焊接缺陷检测算法的特征提取器有望取得较高的效果。

3 Welding Defect Data

In the dataset, the defect types are composed of porosity, lacks of fusion, slag, and cracks。The dataset is composed of radiographic testing images taken differently depending on the steel plate, pipe, and pipe size, thus it can be read and evaluated without dividing the weld after learning.

在数据集中，缺陷类型由孔隙、未熔合、熔渣和裂纹组成。数据集由根据钢板、管道和管道尺寸不同而拍摄的射线照相检测图像组成，因此在学习后无需分割焊缝即可读取和评估。

3.1. Pre-Processing

High-definition images degrade the learning rate, and it is difficult to expect good performance with an increase in the number of parameters to learn.

高清晰度图像降低了学习速度，并且随着要学习的参数数量的增加，很难期望良好的性能。

we removed the rest except for the weld and used it as the training data along with the information marked.

我们移除了除焊缝之外的其余部分，并将其与标记的信息一起用作训练数据。

In this study, we segmented the radiographic testing images to fit the weld into Sections 2–5 (Figure 5). The segmented image becomes smaller from 4900 pixels to less than 980 pixels, and the learning rate was reduced from 1.7 s to 0.3 s per epoch.

在这项研究中，我们将射线照相检测图像分成2-5部分(图5)。分割的图像从4900像素变得更小，小于980像素，并且学习速率从每历元1.7秒降低到0.3秒。

The total number of data increased to 341 from the 134 through image segmentation. Of them, there are 321 training data and 20 validation data.

通过图像分割，数据总量从134个增加到341个。其中，有321个训练数据和20个验证数据。

3.2. Small Object in Faster R-CNN

In MS COCO object detection competition [14], small , medium, and large are classified according to the pixel area, so the welds belong to both small and large object.

在MS COCO对象检测竞赛[14]中，根据像素面积对小型、中型和大型进行分类，因此焊缝同时属于小型和大型对象。

To theoretically estimate the size of the anchor box generated in RPN, we selected the size and aspect ratio by taking into account intersection over union (IoU) according to [15].

为了从理论上估计RPN中生成的锚框的大小，我们根据[15]通过考虑并集上的交集(IoU)来选择大小和纵横比。

4. Experiments and Results

The detection rate-recall graph of the algorithms for the porosity and LoS is shown in Figure 7.

孔隙率和LoS算法的检测率-召回图如图7所示。

Table 3 shows the result of evaluating the performance of the algorithm.

表3示出了评估算法性能的结果。

When comparing the feature extractor, ResNet showed relatively higher performance than Inception-ResNet V2.This shows that the classification of defect classes is lower than that of other models, but the performance of locations is better.

当比较特征提取器时，ResNet显示出比Inception-ResNet V2相对更高的性能。这表明缺陷类别的分类低于其他模型，但是位置的性能更好。

The porosity has similar training and evaluation data, and has a high overall detection rate.

孔隙度具有相似的训练和评估数据，并且具有高的总体检测率。

5. Conclusions

we set the size of the anchor box and aspect ratio to be suitable for small objects, and set the number of region proposal recommendations through an experiment.

我们设置锚框的大小和纵横比以适合小对象，并通过一个实验设置区域提议推荐的数量。

we compared ResNet and Inception-ResNet V2 with the feature extractor of Faster R-CNN, and proposed ResNet with the highest performance.

我们将ResNet和Inception-ResNet V2与更快的R-CNN的特征提取器进行了比较，并提出了性能最高的ResNet。

In this study, we used data conversion for efficient training and performance improvement.The experimental results show that data conversion could increase the performance by 0.074 in radiographic testing images.but covered specific welding defects to increase practicality.

在这项研究中，我们使用数据转换进行有效的培训和绩效改进。实验结果表明，数据转换可以将射线检测图像的性能提高0.074倍。但是覆盖了特定的焊接缺陷以增加实用性。

We increased the data of the LoS through data conversion and image segmentation, but could not significantly decrease the biased results.

我们通过数据转换和图像分割增加了LoS的数据，但不能显著减少有偏差的结果。