Application of X-ray defect image detection technology for ship weldsbased on Faster-RCNN
-
摘要: 将目标检测网络Faster-RCNN应用在船舶焊缝X射线缺陷图像检测中，探讨了Faster-RCNN在X射线焊缝缺陷检测中的效果。针对船舶工业中的X射线焊缝图像，首先采用CLAHE方法对焊缝X射线图像进行预处理，并将焊缝中存在的气孔、裂纹、未熔合等5种具有典型特征的缺陷作为识别目标进行标注并对数据进行增强。在目标识别上，采用ResNet-50作为主干网络来减少梯度弥散现象提高模型准确率，并针对焊缝缺陷目标小的特点对RPN网络锚点参数进行改进优化，同时引入FPN网络提取缺陷特征。最后与其他检测算法进行对比，试验结果表明，该数据集在模型上的mAP值达到96.33%，可以满足X射线焊缝缺陷自动化辅助检测要求。
-
关键词:
- Faster-RCNN /
- X射线 /
- CLAHE /
- 焊缝缺陷检测
Abstract: In this paper, the target detection network Faster-RCNN was applied to the X-ray image defect detection of ship welds, and the effect of Faster-RCNN in the X-ray weld defect detection was discussed. Aiming at the X-ray weld image in the shipbuilding industry, this paper first used the CLAHE method to preprocess the weld X-ray image, and took the five types of defects with typical characteristics such as pores, cracks, and LOF in the weld as the identification target annotated and enhanced the data. In object detection, ResNet-50 was used as the backbone network to reduce the gradient dispersion phenomenon and improve the accuracy of the model. The anchor point parameters of the RPN network were improved and optimized for the characteristics of small weld defects. At the same time, the FPN network was introduced to extract the defect features. Finally, a comparative experiment with other detection methods was carried out. The experimental results showed that the mAP value of the data set on the model reached 96.33%, which can meet the requirements of automatic auxiliary detection of X-ray weld defects.-
Keywords:
- Faster-RCNN /
- X-ray /
- CLAHE /
- detection of weld defect
-
-
[1] 褚慧慧. 基于视觉的焊缝质量检测技术研究[D]. 哈尔滨:哈尔滨工程大学, 2017. [2] COLLOBERT R. Natural language processing from scratch[J]. Journal of Machine Learning Research, 2011(8):2493-2537.
[3] REDMON J, DIVVALA S, GIRSHICK R, et al. You only look once:unified, real-time object detection[C]//Proceedings of the IEEE conference on computer vision and pattern recognition. Las Vegas:IEEE, 2016:779-788.
[4] REDMON J, FARHADI A. YOLO9000:better, faster, stronger[C]//Proceedings of the IEEE conference on computer vision and pattern recognition.Hawaii:IEEE, 2017:6517-6525.
[5] LIU W, ANGUELOV D, ERHAN D, et al. SSD:Single shot multibox detector[C]//Proceedings of European conference on computer vision. Cham:Springer, 2016:21-37.
[6] REN S Q, HE K M, GIRSHICK R, et al.Faster R-CNN:towards real-time object detection with region proposal networks[J].IEEE Transactions on Pattern Analysis and Machine Intelligence, 2017, 39(6):1137-1149.
[7] 胡晓轩, 甄希金, 朱琦, 等. 深度迁移学习下的船舶焊接表面缺陷智能检测系统[J]. 造船技术, 2021, 49(3):84-88. [8] 李砚峰, 刘翠荣, 吴志生, 等.基于深度学习One-stage方法的焊缝缺陷智能识别研究[J]. 广西大学学报(自然科学版), 2021, 46(2):362-372. [9] 郑林涛, 董永生, 史恒亮. 一种新型X射线安检图像增强算法[J]. 科学技术与工程, 2014, 14(23):252-256.
计量
- 文章访问数: 14
- HTML全文浏览量: 0
- PDF下载量: 8
下载:
