高精度地圖資料蒐集處理 與分類評估

根據相關研究指出，自駕車系統主要由三大技 術所組成：首先是對環境的感知能力，讓自駕車攫 取並了解周遭的環境資訊，以利決策如障礙物規避、

車速調整、車道變換及行駛方向判定等；再者為定 位與測繪(Mapping)技術，讓駕駛隨時掌握車輛於地 圖上的位置，並建構周遭三維環境資訊，以利車輛 執行導航決策及更新高精度地圖；最後為行駛決策 (Driving Policy)，依據自駕車所在之場域給予適當 的決策及應對措施(De Silva et al., 2018)。由於自駕 車本身缺乏自我判斷與決策的能力，故無論是自駕 車本身的定位精度，或是圖資上的屬性資料都需要 被精確定義，再者賦予自駕車感知外界環境變化之 能力亦為核心技術之一。

自駕車面臨的其中一項挑戰即為本身絕對位 置的定位精度，傳統 GNSS 定位精度僅能達製公尺 等級，雖然 RTK 等相對定位技術有機會改正至 cm 等級之定位精度，然而於目前環境複雜的都市 地區，多數無法接四顆以上或良好的衛星觀測資

料，因此無法準確將自駕車定位於正確的車道位 置，從而有文獻指出：利用車道辨識(lane detection) 之演算法能有效提升車輛本身之定位精度，協助 車輛於都市地區達到公尺等級的定位要求(Gu et al., 2015) ； 荷 蘭 導 航 圖 資 公 司 TomTom 提 出 自 主 RoadDNA 技術，使自駕車得以利用自身感測器所 得動態地圖對誤差即時修正，以具備穩定且高精 度的定位成果，於此彰顯了地圖輔助之重要性。

為了達成自動駕駛或 ADAS 之安全需求必須 仰賴多感測器間的整合與軟硬體之間的搭配，自 駕車上搭載之標準硬體設備包含相機 、光達、

GNSS、INS、雷達等，可作為車載移動式測繪平 台製作高精度地圖之搭配感測器，其中相機與光 達可以視為車輛的視覺感官系統，透過蒐集的影 像和點雲資料以獲取使用者所處之空間資訊，進 而反饋 INS/GNSS 定向定位，系統致使自駕車對外 界做出安全且正確的定位回應。故可想而知自駕 車會收集來自各個感測器的巨量資料，因此無法 再以傳統人工方式處理，而需藉由機器自動化快 速萃取與更新，並提供使用者正確的決策分

圖 5 自駕車上的感測器系統 (Vardhan, 2017)

根據目前自動駕駛發展趨勢，包含 Google、 Neural Networks, CNN)技術在內，逐步演化成為具 備視覺影像自動化處理技術萃取道路幾何、路標 等高精度地圖可用之特徵物的演算技術，有部分 文獻指出將 CNN 修改成全卷積神經網路(Fully Convolutional Networks, FCN)以辨識影像中的特徵 物，其演算法的優勢在於不受輸入影像尺寸大小 之語意分割(semantic segmentation)與特徵萃取的應 用，如利用基於卷積類神經網路的車道辨識演算 法(lane detection algorithm)，辨識由點雲資料所產 製的影像內的特徵物，以獲取穩定且高精度的辨 識成果(He et al., 2016)；或是利用 SVM(Support Vector Machines)分類演算法從三維點雲資料中辨 識不同型態的交通號誌且能達到 89%辨識精度 (Levinson et al., 2011)；以往機器學習在處理點雲資 料時，點雲資料必須先被轉換成其他表示方式，

如投影成透視影像，在 PointNet(Qi et al., 2017a)、

PointNet++(Qi et al., 2017b) 、 VoxelNet(Zhou and

Tuzel, 2017)這類演算法被提出以前，沒有任何機 器學習是能夠直接以端對端(end-to-end)的架構，來

必須評估建置高精度地圖所具備的環境特徵萃取 Mapping Technology in year 2014, Department of Land Administration, MOI. (in Chinese)]

林耿帆，2012。以物件為基礎之光達點雲分類，國 立臺灣大學土木工程學研究所碩士論文。[Lin, K.F.,2012. Object-based classification for LiDAR point cloud, Master Thesis, National Taiwan University, Taiwan, ROC. (in Chinese)]

Aly, M., 2008, Real time detection of lane markers in urban streets, 2008 IEEE Intelligent Vehicles Symposium, Eindhoven, Netherlands, pp. 7-12.

De Silva, V., Roche, J., and Kondoz, A., 2018. Fusion of LiDAR and camera sensor data for environment sensing in driverless vehicles, arXiv preprint, arXiv:1710.06230.

Gu, Y., Hsu, L.T., and Kamijo, S., 2015. Correction of vehicle positioning error using 3D-map-GNSS and vision-based road marking detection, 2015 IEEE International Conference on Vehicular Electronics and Safety (ICVES), Yokohama, Japan, pp. 140-145.

He, B., Ai, R., Yan, Y., and Lang, X., 2016. Lane marking detection based on Convolution Neural Network from point clouds, 2016 IEEE 19^th International Conference on Intelligent Transportation Systems (ITSC), Rio de Janeiro, Brazil, pp. 2475-2480.

Jiao, J., 2018. Machine learning assisted high-definition map creation, 2018 IEEE 42nd Annual Computer Software and Applications Conference (COMPSAC), Tokyo, Japan, pp.367-373.

Levinson, J., Askeland, J., Becker, J., Dolson, J., Held, D., Kammel, S., Kolter, J.Z., Langer, D., Pink, O., Pratt, V., Sokolsky, M., Stanek, G., Stavens, D., Teichman, A., Werling, M., and Thrun, S., 2011.

Towards fully autonomous driving: Systems and algorithms, 2011 IEEE Intelligent Vehicles Symposium (IV), Baden-Baden, Germany, pp.

163-168.

Long, J., Shelhamer, E., and Darrell, T., 2015. Fully convolutional networks for semantic segmentation, 2015 IEEE Conference on Computer Vision and Pattern Recognition (CVPR), Boston, MA, USA, pp.3431-3440.

NHTSA, 2017. Automated driving systems 2.0: A vision for safety, Available at:

https://www.nhtsa.gov/sites/nhtsa.dot.gov/files/d ocuments/13069a-ads2.0_090617_v9a_tag.pdf, Accessed February 25, 2019.

Qi, C.R., Su, H., Mo, K., and Guibas, L.J., 2017a.

Pointnet: Deep learning on point sets for 3D classification and segmentation, 2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR), Honolulu, HI, USA, pp. 77-85.

Qi, C.R., Yi, L., Su, H., and Guibas, L.J., 2017b.

Pointnet++: Deep hierarchical feature learning on point sets in a metric space, 31st Conference on Neural Information Processing Systems (NIPS 2017), Long Beach, CA ,USA, pp. 5099-5108.

Sasse, V., 2017. NDS & OADF Challenges on Data Necessary to Serve Automated Driving (AD), Available at: http://en.sip-adus.go.jp/evt/workshop2017/file/evt_ws2017_s2 _VolkerSasse.pdf, Accessed February 25, 2019.

Shimada, H., Yamaguchi, A., Takada, H., and Sato, K., 2015. Implementation and evaluation of local dynamic map in safety driving systems, Journal of Transportation Technologies, 5(2): 102-112.

Stephenson, S., Meng, X., Moore, T., Baxendale, A.

and Edwards, T., 2011. Accuracy requirements and benchmarking position solutions for intelligent transportation location based services, 8^th International Symposium on Location-Based Services, Vienna.

Vardhan, H., 2017. HD Maps: New age maps powering autonomous vehicles, Available at:

https://www.geospatialworld.net/article/hd-maps-autonomous-vehicles/, Accessed February 25, 2019.

World Health Organization, 2018. Global status report on road safety, Available at:

https://www.who.int/violence_injury_prevention/

road_safety_status/2018/en/, Accessed February 25, 2019.

Zhou, Y., and Tuzel, O., 2017. Voxelnet: End-to-end learning for point cloud based 3D object detection, arXiv preprint, arXiv:1711.06396.

1Master, Department of Geomatics, National Cheng Kung University Received Date: Feb. 25, 2019

2 Ph.D. Candidate, Department of Geomatics, National Cheng Kung University Revised Date: Sep. 02, 2019

3 Professor, Department of Geomatics, National Cheng Kung University Accepted Date: Sep. 11, 2019

4 Director General, Department of Land Administration, M. O. I.

5 Division Chief, Department of Land Administration, M. O. I.

6 Officer, Department of Land Administration, M. O. I.

* Corresponding Author, Tel: 886-6-2370876 ext.857, E-mail: [email protected]

Feasible Assessment of Regulation and Data Collection Process in