臺灣博碩士論文加值系統

車聯網 (Internet-of-Vehicle; IoV) 的發展與應用，對於改善行車之安 全及品質是可預期的。藉由車輛與車輛 (V2V) 以及車輛與基礎設施 (V2I) 的通訊，為駕駛人提供更完整並且無死角的路況資訊。為了確保 車輛間傳遞的訊息的可靠性，精確的車輛定位則成為了必要的條件， 也是本論文主要探討之研究主題。本文提出一種基於影像來作為車聯 網中車輛定位之系統架構，將行車記錄器所攝錄之影像傳送至路邊設 立之資料庫，並依資料庫中已建立之場景三維點雲模型回傳至車輛做 定位。為了降低資料庫記憶體成本及通訊開銷，以及解決場景因時間 或天氣的光線變化，本文亦提出了點雲模型壓縮及更新之演算法。

This paper presents a method for ego-positioning with low cost monocular cameras for an IoV (Internet-of-Vehicle) system. To reduce the computational and memory requirements as well as the communication overheads, we formulate the model compression algorithm as a weighted k-cover problem for better preserving model structures. Specifically for real-world vision-based positioning applications, we consider the issues with large scene change and propose a model update algorithm to tackle these problems. A long-term positioning dataset with more than one month, 105 sessions, and 14,167 images is constructed. Based on both local and up-to-date models constructed in our approach, extensive experimental results show that sub-meter positioning accuracy can be achieved, which outperforms existing vision-based algorithms.

口試委員會審定書 i
致謝 ii
中文摘要 iii
Abstract iv
Contents v
List of Figures vii
List of Tables x
1 Introduction 1
2 Related Work 4
2.1 Feature Matching with 3D Models 4
2.2 Model Compression 5
2.3 Long-Term Positioning 6
3 Vision-Based Ego-Positioning 7
3.1 Training Phase 7
3.1.1 Image-Based Modelling 7
3.1.2 Structure Preserving Model Compression 9
3.1.3 Model Update 13
3.2 Ego-Positioning Phase 15
3.2.1 2D-to-3D Image Matching and Localization 16
4 Experiment 17
4.1 Simulation of V2V Tracking with GPS and Vison-Based Positioning 18
4.2 Positioning Evaluation of Single Still Image 19
4.3 Positioning Evaluation of Video Sequence 22
4.4 Long-Term Positioning Dataset 24
4.5 Positioning Evaluation with Model Update 25
5 Conclusions and Suggestions for Further Research 28
5.1 Conclusions 28
5.2 Suggestions for Further Research 28
5.2.1 Augmented Reality 29
Bibliography 31

[1] Gps horizontal position accuracy. http://www.leb.esalq. usp.br/disciplinas/Molin/leb447/Arquivos/GNSS/ ArtigoAcuraciaGPSsemAutor.pdf.
[2] C. Arth, D. Wagner, M. Klopschitz, A. Irschara, and D. Schmalstieg. Wide area localizationonmobilephones. InternationalSymposium on Mixed andAugmented Reality,2009.
[3] S. Cao and N. Snavely. Minimal scene descriptions from structure from motion models. CVPR,2014.
[4] T.Driver. Long-termpredictionof gpsaccuracy: Understandingthe fundamentals. IONGNSSInternationalTechnicalMeetingoftheSatelliteDivision,2007.
[5] R. Hartley and A. Zisserman. Multiple view geometry in computer vision. CambridgeUniversityPress,2004.
[6] A. Irschara, C. Zach, J. Frahm, and H. Bischof. From structure-from-motion point cloudstofastlocationrecognition. IEEE Conference on Computer Vision and PatternRecognition,2009.
[7] E. Johns and G. Z. Yang. Dynamic scene models for incremental, long-term, appearance-basedlocalisation. ICRA,2013.
[8] E.JohnsandG.Z.Yang. Featureco-occurrencemaps: Appearance-basedlocalisationthroughoutthedayfeatureco-occurrencemaps: Appearance-basedlocalisation throughouttheday. ICRA,2013.
[9] R.Kalman. Anewapproachtolinearfilteringandpredictionproblems. Journalof BasicEngineering,82(1):35–45,1960.
[10] Y.Li,N.Snavely,andD.Huttenlocher.Locationrecognitionusingprioritizedfeature matching. EuropeanConferenceonComputerVision,2010.
[11] Y.Li,N.Snavely,D.Huttenlocher,andP.Fua. Worldwideposeestimationusing3d pointclouds. EuropeanConferenceonComputerVision,2012.
[12] H. Lim, S. Sinha, M. Cohen, and M. Uyttendaele. Real-time image-based 6-dof localization in large-scale environments. International Symposium on Mixed and AugmentedReality,2012.
[13] H. Liu, T. Mei, J. Luo, H. Li, and S. Li. Finding perfect rendezvous on the go: Accuratemobilevisuallocalizationanditsapplicationstorouting.ACMMultimedia, 2012.
[14] D. Lowe. Distinctive image features from scale-invariant keypoints. International JournalofComputerVision,60(2):91–110,2004.
[15] S.Middelberg,T.Sattler,O.Untzelmann,andL.Kobbelt.Scalable6-doflocalization onmobiledevices. EuropeanConferenceonComputerVision,2014.
[16] M. Modsching, R. Kramer, and K. Hagen. Field trial on gps accuracy in a medium sizecity: Theinfluenceofbuilt-up. WorkshoponPositioning,NavigationandCommunication,2006.
[17] M.MujaandD.Lowe.Fastapproximatenearestneighborswithautomaticalgorithm configuration. International Conference on Computer Vision Theory and Applications,2009.
[18] H. S. Park, Y. Wang, E. Nurvitadhi, J. C. Hoe, Y. Sheikh, and M. Chen. 3d point cloudreductionusingmixed-integerquadraticprogramming. ComputerVisionand PatternRecognitionWorkshops,2013.
[19] D.Reid.Analgorithmfortrackingmultipletargets.IEEETransactionsonAutomatic Control,1979.
[20] T. Sattler, B. Leibe, and L. Kobbelt. Fast image-based localization using direct 2dto-3dmatching. InternationalConferenceonComputerVision,2011.
[21] T. Sattler, B. Leibe, and L. Kobbelt. Improving image-based localization by active correspondencesearch. EuropeanConferenceonComputerVision,2012.
[22] N.Snavely,S.Seitz,andR.Szeliski. Phototourism: Exploringphotocollectionsin 3d. ACMTransactionsonGraphics,25(3):835–846,2006.
[23] J. Ventura and T. Hollerer. Wide-area scene mapping for mobile visual tracking. InternationalSymposiumonMixedandAugmentedReality,2012.
[24] A. Wendel, A. Irschara, and H. Bischof. Natural landmark-based monocular localizationformavs. InternationalConferenceonRoboticsandAutomation,2011.
[25] C. Wu. Siftgpu: A gpu implementation of scale invaraint feature transform (sift). http://cs.unc.edu/~ccwu/siftgpu,2007.
[26] C.Wu. Towardslinear-timeincrementalstructurefrommotion. 3DV,2013.
[27] C.Wu,S.Agarwal,B.Curless,andS.M.Seitz.Multicorebundleadjustment.CVPR, 2011.
[28] M. Y. Yang and W. Forstner. Plane detection in point cloud data. International ConferenceonMachineControlGuidance,2010.