由2010年國道三號走山事件之後，發現台灣地錨邊坡普遍存在著程度不一的地錨繡蝕問題。為了解地錨邊坡的現況，各單位進行了廣泛地地錨檢測工作，其中包含了地錨內視鏡檢測項目。藉由內視鏡檢測，可以快速地看到地錨錨頭下方鋼絞線的鏽蝕狀況，並對鋼絞線之鏽蝕程度予以分級，但因過往都是以人為主觀判斷為主，結果往往是因人而異。因此，本研究將影像辨識技術導入地錨鋼絞線鏽蝕程度之影像判識上，利用Google提供的Colaboratory公開軟體，撰寫及執行Python和Mask R-CNN神經網路模型，並對地錨檢測專業廠商提供之地錨內視鏡鋼絞線的影像圖片，做實例分割和目標檢測，以研判地錨鋼絞線之鏽蝕程度，並做出A級(最嚴重)到D級(輕微)共4級之分級。因受限於現場內視鏡的影像品質，以及案例資料分布以A、B級多於C、D級的情況，因此以影像判釋鋼絞線鏽蝕分級之初步結果，其平均精確度值 ( mAP )僅達0.7左右，但不同廠商之截圖影像和影像級別的分布，對mAP值也會有影響。

After the catastrophic landslide occurred at National Expressway No. 3 in 2010, the ground anchors on the anchored slopes in Taiwan were found suffered from different extents of corrosion through the island wide slope inspection programs. To evaluate the current stability status of anchored slopes, borescope was adopted to inspect the status of steel strands under the anchor head. Based on the borescope images, the corrosion condition of the steel strands was examined and graded by the inspecting engineers. However, it was concerned that such subjective judgments by the engineers might affect the outcome of corrosion grading. This research is aimed to rationalize the grading process on corroded steel strands using the image recognition technology. The open source software Colaboratory by Google was adopted here to write and execute the Python program and Mask R-CNN neural network model to detect the steel strands from the images captured by borescope. Borescope videos and images provided by different companies were used to train the model to differentiate the degrees of steel strand corrosion. Totally, four grade levels from A (very serious) to D (minor) are used to grade the strand corrosion. Due to the not-so-good quality of images obtained from the field borescope inspection and the unevenly distributed case data (i.e., more A and B grade cases than C and D grade cases), the preliminary results only showed a mean average precision (mAP) of about 0.7 for the borescope image recognition; but the mAP values tended to vary with the data distribution and inspection companies.

[1] Girshick, R., Donahue, J., Darrell, T. and Malik, J. (2014) "Rich feature hierarchies for accurate object detection and semantic segmentation", Proceedings of the 2014 IEEE Conference on Computer Vision and Pattern Recognition ser. CVPR ‘14, (pp. 580-587)
[2] Yani Muhamad , Irawan Budhi and Setiningsih Casi (2019) " Application of Transfer Learning Using Convolutional Neural Network Method for Early Detection of Terry’s Nail", International Conference on Electronics Representation and Algorithm (ICERA 2019)
[3] Girshick R. (2015) "Fast r-cnn", Proceedings of the 2015 IEEE International Conference on Computer Vision (ICCV), (pp. 1440-1448)
[4] Ren S., He K., Girshick R. and Sun J. (2015) "Faster r-cnn: Towards real-time object detection with region proposal networks", Proceedings of the 28th International Conference on Neural Information Processing Systems - Volume 1 ser. NIPS'15, (pp. 91-99)
[5] He K., Girshick R., Gkioxari G. and Dollar P. (2018) "Mask R-CNN", IEEE Transactions on Pattern Analysis and Machine Intelligence
[6] Katsamenis Iason, Protopapadakis Eftychios, Doulamis Anastasios, Doulamis Nikolaos, and Athanasios Voulodimos (2020) "Pixel-Level Corrosion Detection on Metal Constructions by Fusion of Deep Learning Semantic and Contour Segmentation", Advances in Visual Computing (pp.160-169)
[7] Attard Leanne, Debono Carl James, Valentino Gianluca (2019) " Automatic Crack Detection using Mask R-CNN", International Symposium on Image and Signal Processing and Analysis (pp.152-157)
[8] Lin T-Y., Maire M., Belongie S., Hays J., Perona P., Ramanan D (2014)"Microsoft coco: Common objects in context" in Computer Vision - ECCV 2014, Cham:Springer International Publishing,( pp. 740-755)
[9] ASTM G1-03, 2003, Standard Practice for Preparing, Cleaning, and Evaluating Corrosion Test Specimen.
[10] Uhlig, H. H. and Revie R. W. (1985). "Corrosion ang Corrosion Control." Wiley Interscience, N.Y.
[11] Garcia-Garcia A., Orts-Escolano S., Oprea S.O., Villena-Martinez V., and Garcia-Rodriguez J.(2017)” A Review on Deep Learning Techniques Applied to Semantic Segmentation”
[12] Mousa May. (2016). "Corrosion behavior of mild steel immersed in different concentrations of NaCl solutions, " Journal of Sebha University-(Pure and Applied Sciences), 15(1), ,( pp .1-12)
[13] Sabatini, Paul J., Dan G. Pass, and Robert C. Bachus (1999). "Geotechnical Engineering Circular No. 4, Ground Anchors and Anchored Systems." FHWA Publication, No. FHWA-IF-99-015.
[14] Liao, H. J. and Cheng, S. H. (2013). "When Ground Anchors Become Rusty." The 1st Taiwan-Kazakhstan Joint Workshop in Geotechnical Engineering, Taipei, Taiwan.
[15] 何明錦、邱建國 、歐昱辰、蔡立倫、何家維 (2009)，「鋼筋腐蝕對於鋼筋混凝土建築構件耐震性能與生命週期之影響–含腐蝕橫向鋼筋的梁構件」內政部建築研究所協同研究報告，ISBN:978-986-02-1769-8。
[16] 韋昀孜 (2014)，「腐蝕鋼筋混凝土柱耐震行為」，國立台灣科技大學，碩士論文。
[17] 楊賢德、鄭清江、黃立遠、陳仁達，（2014），「台北市人工邊坡地錨護坡設施檢測案例介紹」，技師期刊，第67期，第13-25頁。
[18] 陳皇仁 (2016)，「地錨錨頭之功能提升及改善對策」，國立台灣科技大學，碩士論文。
[19] 林秀杰 (2014)，「地錨自由段護管型式對地錨荷重傳遞及周圍地層應力之影響」，國立台灣科技大學，碩士論文。
[20] 廖洪鈞、鄭世豪、陳俊仲(2012)，「台灣地錨設計規範之探討」，地工技術，第132期，頁25-34