高頻率的道路巡檢是現今台灣各級道路維持道路水準及避免因道路毀損而造成傷亡事故.而各級道路之道路巡檢皆有賴於開口合約廠商之巡檢，其提供高頻率之鋪面巡檢以及附屬設施之檢查。主要道路之巡檢頻率從一天至一周為週期依各級機關要求而有所不同，然而廠商巡查之檢測設備以及大量巡修資料卻缺乏後續之相關應用或PCI等數值化之轉換，使巡檢之資料無法達到大數據應用以提升道路長期養護之功效。
因此本篇研究利用檢視現有之道路巡檢方式，並利用現行之巡檢設備做其後端影像辨識軟體之開發，希望能提升現行廠商之巡檢效益，將破壞自動辨別，並導入ASTM D6433-16 之 PCI 數值當中，導出數值化之道路成效。
本篇研究利用行車記錄器以及相關低成本高普及率之影像裝置為主要硬體，利用影片切割至影像以及車行速度之關係來取得完整道路之影像，以SLIC Superpixels 為主要辨識原理所開發，透過兩階段之影像分群已篩選出影像中的鋪面破壞。分群出之破壞再利用破壞分類之判定來界定出補錠、坑洞、縱橫向裂縫、鱷魚狀裂縫，再導入PCI之數值運算。
本篇研究成果與半自動化之鋪面檢測軟體有良好之符合特性，並藉由調整影像截取頻率和行車速度之關係，以達到全面鋪面檢測之目標，成效雖因2D影像之深度量測限制而必須縮減偵測破壞項目，不及傳統之人工巡檢的精細，但卻比現行半自動化更為貼近人工檢測數值，以及可以大幅縮減PCI量測所需耗費之人力及時間成本，未來希望藉由影像裝置之升級以及人工智能學習之深入開發，以提升此套軟體之效益。

High frequency road inspection is the key for keeping the road in good condition all the time. Most of the roads in Taiwan rely on contractor inspection. The contractor inspection is not only offering high frequency inspection but fixing the serious distress on the pavement in no time preventing citizens from accident and casualties.
However, the method and equipment of contractor inspection are not smart enough and the inspection data is hard to apply on further analyze for making the long-term maintenance plan. Therefore, the objective of this study is improving the contractor pavement inspection nowadays through popular and low-cost equipment and developing the software for recognizing pavement distress from the capturing image automatically.
Pavement Condition Index (PCI) could show the pavement performance very well according to the ASTM D6433-16. The conventional method way of the PCI survey is manual inspection which is time consuming and manpower consuming. Therefore, Automatic PCI survey and Semi-Automatic PCI survey are the methods to improve the conventional method.
This study used film from the dash cam and SLIC superpixels method to recognizing the distress. Through two times clustering extract the distress and the length known object in the image to calibrate the image parameter scale. After the previous work, the software would input the distress quantity into the PCI calculation.
The performance of the software has great consistent with the Semi-Auto method but since the 2D image could get the depth of the distress; the result of the conventional manual inspection has better performance.
The software could find out serious distress on the road automatically and through the equipment upgrade and the machine-learning technique could upgrade the software in the future.

Reference
1. ASTM D6433-99(1999) Standard Practice of Roads and Parking Lots Pavement Condition Index Surveys.
2. Construction and planning agency ministry of the interior，“ Urban road maintenance and management spec ” , 2003。
3. ASTM D6433-16(2016) Standard Practice of Roads and Parking Lots Pavement Condition Index Surveys.
4. SHRP, “Distress Identification Manual for the Long-Tern Pavement Performance Project”, Washington, D.C., 2014.
5. Tsung-Hsun Sung, “The study of surveying asphalt pavement condition index”, National Central University, Unpublished master's thesis, 2004.
6. Kun-Hu Lin, “Establishment of the urban road network level pavement management structure - A Case Study in Taipei”, National Central University, Unpublished doctor's thesis, 2015.
7. New Construction office, Public Works Department, Taipei City Government Website:
http://nco.gov.taipei
8. Directorate General of Highways， Highway management and maintenance information system , 2017.
9. New Taipei City, The Grievance Hotline for Road Maintenance APP ,2017。
10. Chang-Ching Lee, “A Study of Applying Line Scan CCD Camera on Pavement Distress Image Survey”, National Central University, Unpublished master's thesis, 2005.
11. Chin-Yuan Zheng, “Auto Pavement Damage Image Detection System Importing Pavement Distress Maintenance Management System”, National Central University, Unpublished master's thesis, 2012.
12. Hou-Yi Chen, “The Study of Image Localization Technology Applied for Inspections Operations on National Expressway”, National Central University, Unpublished master's thesis, 2012.
13. Lin-Jyun Chen ,“ A Study of Image Recognition Technology Applying to Pavement Distress”, National Central University, Unpublished master's thesis, 2013.
14. Zhi-Sheng Xue, “ Application of Image Recognition Study into Pavement Distress 3D-model Construction Potholes”, National Central University, Unpublished master's thesis, 2015.
15. NichirekiWebsite2017/03/05:https://www.nichireki.co.jp/english/product/consult/consult_list05/consult0505.html
16. Taipei Citity Government and National Central University, “Taipei Urban Road Long-Term Life Cycle Surveying and Maintenance Plan” (2017).
17. Mulry, B., Jordan, M., O'Brain, D., “Automated Pavement Condition Assessment Using Laser Crack Measurement System (LCMS) on Airfield Pavements in Ireland”,9th International Conference on Managing Pavement Assets (2015)
18. Goozalez, R., C., “Digital Image Processing 4th”, Person, (2007).
19. Miraliakbari, A., Sok, S., Ouma, Y. O, Hahn, M., “Comparative Evaluation of Pavement Crack Detection Using Kernel-Based Technique in Asphalt Road Surfaces”
20. Gavilán, M., Balcones, D., Marcos, O., Llorca, D.F., Sotelo, M.A., Parra, I., Ocaña, M., Aliseda, P., Yarza, P., Amírola, “Adaptive Road Crack Detection System by Pavement Classification”, Sensors, November, 9628-9657 (2011).
21. Achanta, R., Shaji, A., Smith, K., Lucchi, A., Fua, P., Süsstrunk, S., “SLIC Superpixels Comparedto State-of-the-Art Superpixel Methods”, IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE, Vol.34, No.11, November (2012)
22. Ying, L. and Salari, E., “Beamlet Transform-Based Technique for Pavement Crack Detection and Classification”, Computer-Aided Civil and Infrastructure Engineering 25, 572-280, (2010).
23. MathWorksWebsite2017/02/05:http://www.mathworks.com/help/vision/ug/camera-calibration.html?requestedDomain=www.mathworks.com
24. Peggy, S., Jean, D., Vincent, L., Dominique, B., “Automation of Pavement Surface Crack Detection Using the Continous Wavelet Transform”, ICIP, September (2006).
25. Donoho, D.L., and Huo, X., “Beamlets and Multiscale Image Analysis” Stanford University and Georgia Institute of Technology, September (2001).
26. Pavement Services, Inc., “Pavement Condition Index Survey &Evaluation of the City of Veneta’s Street Network”, Veneta, February (2015).
27. Ching-Chun Lu , “ Application of Image Recognition Study into Cracking on Bridge ”, National Central University, Unpublished master's thesis, 2014.
28. Chine-Ta Chen, “ Study of Automatic Pavement Roughness Measurement and Distress Image Detection System”, National Central University, Unpublished doctoral thesis, 2009.
29. Fankhauser, P., Bloesch, M., Rodriguez,D., Kaestner, R., Hutter, M., Siegwart, R., “Kinect v2 for Mobile Robot Navigation: Evaluation and Modeling” International Conference on Advanced Robotics(ICAR), Istanbul, July (2015).
30. New Construction office, Public Works Department, Taipei City Government Website:
http://nco.gov.taipei
31. Mertz, C., Varadharajan, S., Jose, S., Sharma, K., Wander, L., Wang, J., “City-Wide Road Distress Montoring with Smartphones” Proceedings of ITS World Congress, September (2014).
32. Byoung Jik Lee, “Position-Invariant Neural Network for Digital Pavement Crack Analysis ”, Computer-Aided Civil and Infrastructure Engineering 19, 105-118, (2004).