結論與未來展望

5.1 結論

本研究期望使用類神經網路來進行閥栓巡檢照片檢查系統以及道路破 損檢測系統，為此本論文先以YOLOv3-tiny 類神經網路進行閥栓巡檢系統之 模型建置，並且在嘗試多種類的近遠照混和以及增加負面訓練資料集等不同 的條件 後， 確認 限 縮在近 照之 內的 閥 栓辨識 模型 效果 最 佳。此 模型 之 Precision 可以到達 99.23%，Recall 也有 98.84%。而在多了地上式消防栓的 種類之後Precision 變為 98.41%，Recall 為 97.85%，都僅有微幅降低。且其 辨識效果並不受照片光影變化以及照片方向性之問題影響。 降低些許 Precision 的前提下提升模型 Recall。依據此實驗結果證明，使用 YOLOv3-tiny 可以建構出準確的閥栓辨識模型與門牌辨識模型，且此閥栓辨 能有效提升Recall，並且微幅提升 Precision。並從實驗結果來看，YOLO 網

路在道路破損檢測上，對於坑洞(Pothole)以及龜裂(Alligator Crack)的偵測效 果是比較好的。與之前使用Single Shot Detection 之文獻[23]相比，此篇文獻 中坑洞的Recall 僅有 2%，而以本研究中所使用之 YOLOv3 網路以及本研究 中所設定之參數，其Recall 高上許多。但其餘標線破損等辨識是劣於此篇論 文的，這一點仍須改進。

5.2 未來展望

閥栓巡檢系統建置所需之模型已經獲得 99.23%之準確率以及 98.84%之 召回率，未來期望能將此一模型配套上資料庫，實現「實時閥栓照片檢測系 統」，在作業人員上傳照片之同時，即可提醒所拍攝照片是否合格以及照片 上傳順序是否有誤，用以減少照片出錯使得外業人員重複進行作業以及內業 作 業 之 成 本 。 未 來 也 希 望 能 將 門 牌 辨 識 配 合 上 Optical Character Recognition(OCR，光學文字辨識)，以期能在照片上傳之同時，即可完成地 址之辨識。

而使用類神經網路進行道路破損檢測此一研究尚未有研究團隊有突破 性的進展，而使用整張照片進行辨識的團隊也不多。在目前目標檢測的演算 法尚在持續發展中，未來希望能使用其餘神經網路架構如 CornerNet [26]進 行辨識，以及使用Octave Convolution [27]替換掉原本 YOLO 中的卷積層，

以期能提升道路破損檢測之效果。

參考文獻

[3] A. Mednis, G. Strazdins, R. Zviedris, G. Kanonirs, and L. Selavo, “Real time pothole detection using android smartphones with accelerometers,”

2011 International conference on distributed computing in sensor systems and workshops (DCOSS), IEEE, pp. 1-6, Jun. 2011.

[4] I. Abdel-Qader, O. Abudayyeh, and M. E. Kelly, “Analysis of edge-detection techniques for crack identification in bridges,” Journal of Computing in Civil Engineering, vol. 17, no. 4, pp. 255-263, Oct. 2003.

[5] Y. Shi, L. Cui, Z. Qi, F. Meng, and Z. Chen, ‘‘Automatic road crack detection using random structured forests,’’ IEEE Transactions on Intelligent Transportation Systems, vol. 17, no. 12, pp. 3434–3445, Dec.

2016.

[6] H. Oliveira and P. L. Correia, ‘‘Automatic road crack detection and characterization,’’ IEEE Transactions on Intelligent Transportation Systems, vol. 14, no. 1, pp. 155–168, Mar. 2013.

[7] L. Zhang, F. Yang, Y. D. Zhang, and Y. J. Zhu, ‘‘Road crack detection using deep convolutional neural network,’’ 2016 IEEE international conference on image processing (ICIP), IEEE, pp. 3708-3712, Sep. 2016.

[8] Y.-J. Cha, W. Choi, and O. Buy¨ uk¨ ozt ¨ urk, “Deep learning-based crack ¨ damage detection using convolutional neural networks,”

Computer‐Aided Civil and Infrastructure Engineering, vol. 32, no. 5, pp.

361–378, Mar. 2017.

[9] 新北市路平報馬仔，https://rdm.ntpc.gov.tw/Road/NewCase.aspx [10] labelImg https://github.com/tzutalin/labelImg

[11] J. Redmon, "Darknet: Open Source Neural Networks in C,"

http://pjreddie.com/darknet/, 2013~2016.

[12] R. Girshick, J. Donahue, T. Darrell, and J. Malik, “Rich feature hierarchies for accurate object detection and semantic segmentation,”

Proceedings of the IEEE conference on computer vision and pattern recognition, pp. 580-587, Jun. 2014.

[13] K. He, X. Zhang, S. Ren, and J. Sun, “Spatial pyramid pooling in deep convolutional networks for visual recognition,” IEEE transactions on pattern analysis and machine intelligence, vol. 37, no. 9, pp. 1904–1916, Sep. 2015.

[14] R. Girshick, “Fast r-cnn,” Proceedings of the IEEE international conference on computer vision, pp. 1440-1448, 2015.

[15] S. Ren, K. He, R. Girshick, and J. Sun, “Faster R-CNN: Towards real-time object detection with region proposal networks,” Advances in neural information processing systems, pp. 91-99, 2015.

[16] J. Redmon, S. Divvala, R. Girshick, and A. Farhadi, “You only look once:

Unified, real-time object detection,” Proceedings of the IEEE conference on computer vision and pattern recognition, pp. 779-788, 2016.

[17] J. Redmon and A. Farhadi, “YOLO9000: better, faster, stronger,”

Proceedings of the IEEE conference on computer vision and pattern recognition, pp. 7263-7271, 2017.

[18] J. Redmon and A. Farhadi, “Yolov3: An incremental improvement,”

2018, arXiv:1804.02767.

[19] W. Liu, D. Anguelov, D. Erhan, C. Szegedy, and S. Reed, “SSD: Single shot multibox detector,” arXiv:1512.02325, 2015.

[20] H. Oliveira and P. L. Correia, ‘‘Automatic road crack segmentation using entropy and image dynamic thresholding,’’ 2009 17th European Signal Processing Conference, IEEE, pp. 622-626, Aug. 2009.

[21] H. Ma, N. Lu, L. Ge, Q. Li, X. You, and X. Li, “Automatic road damage detection using high-resolution satellite images and road maps,”

Geoscience and Remote Sensing Symposium (IGARSS), 2013 IEEE International, pp. 3718–3721, 2013.

[22] F.-C. Chen and M. R. Jahanshahi, "NB-CNN: Deep Learning-Based Crack Detection Using Convolutional Neural Network and Naïve Bayes Data Fusion," IEEE Transactions on Industrial Electronics, vol. 65, no.

5, pp. 4392-4400, 2018.

[23] H. Maeda, Y. Sekimoto, T. Seto, T. Kashiyama, and H. Omata, “Road damage detection and classification using deep neural networks with smartphone images,” Computer‐Aided Civil and Infrastructure Engineering , vol. 33, no. 12, pp. 1127-1141, 2018.

[24] T.-Y. Lin, P. Dollár, R. Girshick, K. He, B. Hariharan, and S. Belongie,

“Feature pyramid networks for object detection,” Proceedings of the IEEE conference on computer vision and pattern recognition, pp. 2117-2125, 2017.

[25] K. He, X. Zhang, S. Ren, and J. Sun, “Deep residual learning for image recognition,” Proceedings of the IEEE conference on computer vision and pattern recognition, pp. 770-778, 2016.

[26] H. Law and J. Deng, “Cornernet: Detecting objects as paired keypoints,”

Proceedings of the European Conference on Computer Vision (ECCV), pp. 734-750, 2018.

[27] Y. Chen, H. Fang, B. Xu, Z. Yan, Y. Kalantidis et al., “Drop an octave:

Reducing spatial redundancy in convolutional neural networks with octave convolution,” arXiv:1904.05049, 2019.