-
摘要: 針對熱軋帶鋼表面缺陷檢測存在的智能化水平低、檢測精度低和檢測速度慢等問題, 本文提出了一種基于自適應全局定位網絡(Adaptive global localization network, AGLNet)的深度學習缺陷檢測算法. 首先, 引入了一種殘差網絡(Residual network, ResNet)與特征金字塔網絡(Feature pyramid network, FPN)集成的特征提取結構, 減少缺陷語義信息在層級傳遞間的消失; 其次, 提出基于Tree-structure parzen estimation的自適應樹型候選框提取網絡(Adaptive tree-structure region proposal network, AT-RPN), 無需先驗知識的測試積累, 避免了人為調參的訓練模; 最后, 引入了全局定位算法(Global localization regression)算法以全局定位的模式在復雜的缺陷檢測中實現缺陷更精確定位.本文實現一種快速、準確、更智能化、更適用于實際工業應用的熱軋帶鋼表面缺陷的算法.實驗結果表明, AGLNet在NEU-DET熱軋帶鋼表面缺陷數據集上的檢測速度保持在11.8fps, 平均精度達到了79.90 %, 優于目前其他深度學習帶鋼表面缺陷檢測算法; 另外該算法還具備較強的泛化能力.
-
關鍵詞:
- 表面缺陷檢測 /
- 深度學習 /
- 特征金字塔 /
- 自適應樹形候選框提取 /
- 全局定位
Abstract: A deep learning defect detection model based on Adaptive global localization network (AGLNet) is presented to solve the problems of low intelligence, low detection accuracy and slow detection speed in Hot-rolled strip surface defect detection. First, the feature extraction structure is combined with Residual network (ResNet) and Feature pyramid network (FPN) to reduce the disappearance of defect semantic information between layers transfers. Secondly, an Adaptive tree-structure region proposal network (AT-RPN) based on Tree-structure parzen estimation algorithm is proposed, which does not need the accumulation of prior knowledge, and avoids the training model by manual parameter adjustment. Finally, a Global localization regression algorithm is proposed to locate defects more accurately in complex defect detection using global positioning mode. In this paper, a fast, accurate, more intelligent algorithm for surface defects detection of hot-rolled strips is realized. The experimental results show that the detection speed of AGLNet remains 11.8 fps and the average accuracy is 79.90 %, which is better than other deep learning algorithms for strip surface defect detection on NEU-DET dataset. In addition, the algorithm has a strong generalization ability. -
表 1 AGLNet、Gird R-CNN and Faster R-CNN基于NEU-DET數據集的對比測試結果
Table 1 Comparison results of AGLNet, Gird R-CNN and Fast R-CNN based on NEU-DET dataset
裂紋 夾雜 斑塊 麻點 壓入氧化 劃痕 AGLNet Grid R-CNN Faster R-CNN 表 2 各個模型在NEU-DET數據集的缺陷檢測平均精度結果(%)
Table 2 Average accuracy results of defect detection on NEU-DET dataset of comparative experiment (%)
方法 平均精度均值 裂紋 夾雜 斑塊 麻點 壓入氧化 劃痕 Faster R-CNN 79.20 71.31 84.63 82.92 80.17 80.31 75.87 RetinaNet 75.36 53.02 78.74 93.33 91.37 62.21 73.49 FCOS 75.18 52.41 75.03 91.48 84.85 62.86 84.43 Grid R-CNN 73.14 41.52 78.68 86.23 86.47 59.74 86.21 YOLO-v1 62.90 42.35 63.42 68.23 66.49 69.37 67.53 YOLO-v2 66.53 47.35 70.47 72.23 65.82 65.49 77.84 YOLO-v3 69.40 68.39 61.88 71.44 68.33 72.66 73.71 YOLO-v4 77.99 64.87 70.84 93.24 83.83 69.52 85.63 YOLO-v5 76.82 62.42 75.76 84.23 81.27 64.59 92.63 YOLOF 77.32 63.48 71.82 90.56 85.21 64.24 88.63 AGLNet 79.90 54.72 83.31 88.63 91.67 64.42 96.64 表 3 各模型FLOPs, Params和FPS對比結果
Table 3 Comparison of Flops, Params and FPS of each model
方法 FLOPs Params Fps Faster R-CNN 408.36GMac 98.25 M $\sim$8.2 RetinaNet 239.32GMac 37.74 M $\sim$12.3 FCOS 438.68GMac 89.79 M $\sim$9.3 Grid R-CNN 329.51GMac 64.32 M $\sim$10.2 YOLO-v3 89.45GMac 27.84 M $\sim$15.4 YOLOF 151.47GMac 63.24M $\sim$13.4 AGLNet 273.95GMac 79.8 M $\sim$11.8 表 4 各類缺陷在不同IoU閾值下的測試結果
Table 4 Detection results of various defects under different IOU thresholds
IoU閾值 缺陷類型 gts Dets Recall mAP IoU0.5 裂紋 139 1 886 0.935 54.72 IoU0.75 裂紋 139 1 823 0.923 47.48 IoU0.5 夾雜 181 1 188 0.945 83.31 IoU0.75 夾雜 181 1 163 0.932 82.17 IoU0.5 斑塊 151 627 0.960 88.63 IoU0.75 斑塊 151 591 0.942 89.45 IoU0.5 麻點 88 689 0.955 91.67 IoU0.75 麻點 88 636 0.938 89.24 IoU0.5 壓入氧化 126 1 034 0.893 64.42 IoU0.75 壓入氧化 126 1 051 0.882 59.66 IoU0.5 劃痕 117 317 0.991 96.64 IoU0.75 劃痕 117 322 0.986 92.79 IoU0.5 全部缺陷 802 5 741 0.947 79.90 IoU0.75 全部缺陷 802 5 586 0.934 76.79 表 5 消融實驗結果
Table 5 Ablation results
序號 ResNet50+FPN ResNet50 AT-RPN RPN mAP(%) fps GPU Memory Usage 1 √ √ 79.90 11.8 5568MiB 2 √ √ 78.64 10.3 7039MiB 3 √ √ 77.97 12.2 5024MiB 4 √ √ 76.82 10.6 6436MiB 表 6 消融實驗對比結果
Table 6 Comparison results of ablation experiments
序號 對比試驗 mAP提升 fps提升 節約顯存占用率 1 實驗1*實驗2 1.26% 1.5 20.89% 2 實驗3*實驗4 1.15% 1.6 21.93% 3 實驗1*實驗3 1.93% -0.4 -10.82% 4 實驗2*實驗4 1.82% -0.3 -9.36% 5 實驗1*實驗4 3.08% 1.2 13.49% 表 7 PCB-Master數據集基本信息
Table 7 Basic information of PCB master dataset
缺陷類型 圖像數量 缺陷數量 漏孔(Missing_hole) 115 497 鼠咬(Mouse_bite) 115 492 斷路(Open_circuit) 115 482 短路(Short) 115 491 毛刺(Spur) 115 488 表 8 各個模型在PCB-Master數據集上測試結果(%)
Table 8 Test results of each model on PCB master dataset (%)
Faster R-CNN RetinaNet FCOS Grid R-CNN Yolo-v3 YOLOF AGLNet MAP(%) 86.9 91.16 88.9 94.3 79.8 95.0 96.9 漏孔 0.874 0.915 0.907 0.956 0.858 0.942 0.995 鼠咬 0.849 0.905 0.852 0.934 0.793 0.934 0.952 斷路 0.862 0.897 0.847 0.915 0.747 0.886 0.929 短路 0.895 0.922 0.928 0.997 0.832 0.997 0.997 毛刺 0.869 0.953 0.915 0.954 0.826 0.989 0.997 余銅 0.865 0.875 0.880 0.905 0.731 0.954 0.942 表 9 PCB-Master測試集檢測數據統計
Table 9 Data statistics of PCB-Master defect detection test set
缺陷類別 gts Dets recall ap 漏孔(Missing_hole) 169 696 0.998 0.995 鼠咬(Mouse_bite) 142 665 0.990 0.952 斷路(Open_circuit) 142 667 0.990 0.929 短路(Short) 132 590 1 0.997 毛刺(Spur) 143 687 1 0.997 余銅(Spurious_copper) 137 644 0.979 0.942 全部缺陷總計 865 3949 亚洲第一网址_国产国产人精品视频69_久久久久精品视频_国产精品第九页 -
[1] 王典洪,甘勝豐,張偉民,雷維新. 基于監督雙限制連接Isomap算法的帶鋼表面缺陷圖像分類方法. 自動化學報, 2014, 40(5): 883-891Wang D H, Gan S F, Zhang W M, Lei W X. Strip Surface Defect Image Classification Based on Double-limited and Supervised-connect Isomap Algorithm. Acta Automatica Sinica, 2014, 40(5): 883-891 [2] Song, K C and Yan Y H. A noise robust method based on completed local binary patterns for hot-rolled steel strip surface defects. Applied Surface Science, 2013, 285(x): 858-864 [3] Neogi, N, Mohanta, D K, Dutta, P K. Review of vision-based steel surface inspection systems. EURASIP Journal on Image and Video Processing, 2014, 1-19 [4] 許志祥,盧宏,沈劍. 攝像機定標及其誤差分析. 自動化學報, 1993, 01(x): 115-117 doi: 10.16383/j.aas.1993.01.018Xu Z X, Lu H, Shen J.Camera Calibration and its Error Analysis. Acta Automatica Sinica, 1993, 01(x): 115-117 doi: 10.16383/j.aas.1993.01.018 [5] 李少波,楊靜,王錚,朱書德,楊觀賜. 缺陷檢測技術的發展與應用研究綜述. 自動化學報, 2020, 46(11): 2319-2336 doi: 10.16383/j.aas.c180538Li S B, Yang J, Wang Z, Zhu S D, Yang G C. Review of development and application of defect detection technology. Acta Automatica Sinica, 2020, 46(11): 2319-2336 doi: 10.16383/j.aas.c180538 [6] 劉國梁,余建波. 基于堆疊降噪自編碼器的神經-符號模型及在晶圓表面缺陷識別. 自動化學報, 2021, x(x): 1-15Liu G L, Yu J B. Application of neural-symbol model based on stacked denoising auto-encoders in wafer map defect recognition. Acta Automatica Sinica, 2021, x(x): 1-15 [7] Ren S, He K, Girshick R B, Sun J. Faster R-CNN: Towards Real-Time Object Detection with Region Proposal Networks. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2015, 39(x): 1137-1149 [8] He K, Gkioxari G, Dollár P Girshick R B. Mask R-CNN. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2020, 42(x): 386-397. [9] Redmon J, Divvala S, Girshick R B, Farhadi A. You Only Look Once: Unified, Real-Time Object Detection. In: 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR). 2016: 779?788 [10] Liu W, Anguelov D, Erhan D, Szegedy C, Reed S, Fu C, Berg A. SSD: Single Shot MultiBox Detector. In: 2016 Proceedings of the European Conference on Computer Vision (ECCV). 2016 [11] Lin T, Dollár P, Girshick R B, He K, Hariharan B, Belongie S J. Feature Pyramid Networks for Object Detection. In: 2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR). 2017: 936?944 [12] Tao X, Zhang D, Wang Z, Liu X, Zhang H, Xu D. Detection of Power Line Insulator Defects Using Aerial Images Analyzed With Convolutional Neural Networks. IEEE Transactions on Systems, Man, and Cybernetics: Systems, 2020, 50(x): 1486-1498 [13] He Y, Song K, Meng Q, Yan Y. An End-to-End Steel Surface Defect Detection Approach via Fusing Multiple Hierarchical Features. IEEE Transactions on Instrumentation and Measurement, 2020, 69(x): 1493-1504 [14] Cheng X, Yu J. RetinaNet With Difference Channel Attention and Adaptively Spatial Feature Fusion for Steel Surface Defect Detection. IEEE Transactions on Instrumentation and Measurement, 2021, 70(x): 1-11 [15] Chen J, Liu Z, Wang H, Nú?ez A, Han Z. Automatic Defect Detection of Fasteners on the Catenary Support Device Using Deep Convolutional Neural Network. IEEE Transactions on Instrumentation and Measurement, 2018, 67(x): 257-269 [16] Zhang C B, Chang C C, Jamshidi M. Concrete bridge surface damage detection using a single-stage detector. IComputer-Aided Civil and Infrastructure Engineering, 2020, 35(x): 389-409 [17] Zhang S, Chi C, Yao Y, Lei Z, Li S. Bridging the Gap Between Anchor-Based and Anchor-Free Detection via Adaptive Training Sample Selection. In: 2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). 2020: 9756?9765 [18] Law H, Deng J. CornerNet: Detecting Objects as Paired Keypoints. International Journal of Computer Vision, 2019, 128(x): 642-656 [19] Duan K, Bai S, Xie L, Qi H, Huang Q, Tian Q. CenterNet: Keypoint Triplets for Object Detection. In: 2019 IEEE/CVF International Conference on Computer Vision (ICCV). 2019: 6568?6577 [20] Jia X, Yang X, Yu X, Gao H. A Modified CenterNet for Crack Detection of Sanitary Ceramics. In: IECON 2020 The 46th Annual Conference of the IEEE Industrial Electronics Society. 2020: 5311?5316 [21] Zhu C, He Y, Savvides M. Feature Selective Anchor-Free Module for Single-Shot Object Detection. In: 2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). 2019: 840?849 [22] Tian Z, Shen C, Chen H, He T. FCOS: Fully Convolutional One-Stage Object Detection. In: 2019 IEEE/CVF International Conference on Computer Vision (ICCV). 2019: 9626-9635. [23] Kong T, Sun F, Liu H, Jiang Y, Li L, Shi J. FoveaBox: Beyound Anchor-Based Object Detection. IEEE Transactions on Image Processing, 2019, 29(x): 7389-7398 [24] Lu X, Li B, Yue Y, Li Q, Yan J. Grid R-CNN. In: 2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). 2019: 7355?7364 [25] Wang J, Zhang W, Cao Y, Chen, K, Pang J, Gong T, Shi J, Loy C C, Lin D. Side-Aware Boundary Localization for More Precise Object Detection. In: 2020 Proceedings of the European Conference on Computer Vision(ECCV). 2020 [26] He Y, Song K, Meng Q, Yan Y. An End-to-End Steel Surface Defect Detection Approach via Fusing Multiple Hierarchical Features. IEEE Transactions on Instrumentation and Measurement, 2019, 69(x): 1493-1504 [27] Bergstra J, Bardenet R, Bengio Y, Kégl B. Algorithms for Hyper-Parameter Optimization. NIPS. 2011 [28] Cao J, Cholakkal H, Anwer R M, Khan F, Pang Y, Shao L. D2Det: Towards High Quality Object Detection and Instance Segmentation. In: 2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). 2020: 11482?11491 [29] Ding R, Dai L, Li G, Liu H. TDD-net: a tiny defect detection network for printed circuit boards. CAAI Trans. Intell. Technol, 2019, 4(x): 110-116
計量
- 文章訪問數: 507
- HTML全文瀏覽量: 200
- 被引次數: 0