Research on defect detection of X-ray weld image based on sparse description
-
摘要: 针对X射线焊缝图像缺陷检测的准确率问题,提出运用Log-Polar变换的距离不变性和角度不变性将缺陷的位置及形状转化为典型缺陷图像的简单二维平面的平移,解决了缺陷和疑似缺陷区域的标定问题。此外,为了提高缺陷识别的检出率及识别准确性,提出基于稀疏描述的缺陷识别,运用从海量数据中提取典型样本、非参数化模型构建以及基于最优方向法的稀疏解求解三大知识体系,对所标定SDR进行识别。实验得出,通过有限的样本训练所得字典矩阵对缺陷的识别率达到了98.5%以上。
-
关键词:
- 缺陷检测 /
- Log-Polar变换 /
- 稀疏描述 /
- 最优方向法
Abstract: Aiming at the accuracy of X-ray weld image defect detection,this paper proposes to use the distance invariance and angle invariance of Log-Polar transformation to transform the position and shape of the defect into the translation of a simple two-dimensional plane of typical defect images, and solve the defects and suspected defects. area calibration problem. In addition, in order to improve the detection rate and recognition accuracy of defect identification, a defect identification based on sparse description is proposed, which uses three major knowledges: extracting typical samples from massive data, building non-parametric models, and solving sparse solutions based on the optimal direction method. system to identify the calibrated SDR. Experiments show that the recognition rate of the dictionary matrix obtained by limited sample training has reached more than 98.5%.-
Key words:
- defect detection /
- Log-Polar transform /
- sparse description /
- optimal direction method
-
表 1 焊缝图像信息表
图像ID 分辨率 位深度 数据量 001-200 2 412×648 24 200 201-400 5 684×531 8 200 表 2 2 412×648实验结果
实验编号 缺陷样本数量 人为标定底片数量 测试底片数量 缺陷识别率/(%) 噪声识别率/(%) 1 8 5 195 98.1 100 2 17 10 190 98.3 100 3 29 15 185 98.3 100 4 36 20 180 98.5 100 5 48 25 175 98.5 100 6 58 30 170 98.8 100 7 66 35 165 98.8 100 8 75 40 160 99.0 100 9 89 45 155 99.2 100 10 93 50 150 99.2 100 11 10 5 195 98.0 100 12 20 10 190 98.2 100 13 31 15 185 98.2 100 14 38 20 180 98.5 100 15 46 25 175 98.5 100 16 55 30 170 98.8 100 17 64 35 165 99.0 100 18 76 40 160 99.2 100 19 88 45 155 99.2 100 20 97 50 150 99.4 100 表 3 2 412×648实验结果
实验编号 缺陷样本数量 人为标定底片数量 测试底片数量 缺陷识别率/(%) 噪声识别率/(%) 21 6 3 197 98.4 100 22 9 6 194 98.5 100 23 16 8 192 98.7 100 24 18 11 189 98.7 100 25 22 12 188 98.9 100 26 27 15 185 99.0 100 27 31 17 183 99.2 100 28 36 19 181 99.2 100 29 43 21 179 99.4 100 30 46 23 177 99.6 100 31 7 4 196 98.3 100 32 11 7 193 98.5 100 33 17 11 189 98.5 100 34 20 13 187 98.7 100 35 23 16 184 98.7 100 36 29 18 182 99.2 100 37 32 21 179 99.3 100 38 35 23 177 99.5 100 39 41 25 175 99.5 100 40 44 28 172 99.7 100 -
[1] 焦敬品, 李思源, 常予, 等. 集箱管接头内焊缝表面缺陷识别方法研究[J]. 仪器仪表学报, 2017(12): 3044-3052. [2] Provencal E, Laperrière L. Identification of weld geometry from ultrasound scan data using deep learning[J]. Procedia CIRP, 2021, 104: 122-127. doi: 10.1016/j.procir.2021.11.021 [3] Patil R V, Reddy Y P. Weld imperfection classification by texture features extraction and local binary pattern[M]. Modeling, Simulation and Optimization. Springer, Singapore, 2021: 367-378. [4] Yan X L, Xu W, Yang G L, et al. Star map recognition algorithm based on improved log-polar transformation[J]. Acta Optica Sinica, 2021, 41(10): 1010001. doi: 10.3788/AOS202141.1010001 [5] Saha S, Rahaman G M A, Islam T, et al. Retinal image registration using log-polar transform and robust description of bifurcation points[J]. Biomedical Signal Processing and Control, 2021, 66: 102424. doi: 10.1016/j.bspc.2021.102424 [6] Sato Y, Bao Y. High speed and high precision facial recognition using log-polar transformation[C]. 2018 International Conference on Sensor Networks and Signal Processing (SNSP). IEEE, 2018: 375-381. [7] Picariello E, Balestrieri E, Picariello F, et al. A new method for dictionary matrix optimization in ECG compressed sensing[C]. 2020 IEEE International Symposium on Medical Measurements and Applications (MeMeA). IEEE, 2020: 1-6. [8] 王雪, 隋立春, 杨振胤, 等. 最优方向耦合字典学习的遥感影像超分辨率重建[J]. 计算机工程与应用, 2018(7): 201-205. [9] 王军敏, 芦荐宇, 张诗梦. 主分量分析在图像信号处理中的应用[J]. 平顶山学院学报, 2020(5): 49-53. [10] Ben Mhenni R, Bourguignon S, Ninin J. Global optimization for sparse solution of least squares problems[J]. Optimization Methods and Software, 2021: 1-30. [11] Zhang S, Xu L, Yan S. A low complexity OMP sparse channel estimation algorithm in OFDM system[C]. 2021 IEEE International Conference on Signal Processing, Communications and Computing (ICSPCC). IEEE, 2021: 1-5.