amateur augmented reality based vehicle navigation system
play

Amateur: Augmented Reality Based Vehicle Navigation System Chu Cao 1 - PowerPoint PPT Presentation

Amateur: Augmented Reality Based Vehicle Navigation System Chu Cao 1 , Zhenjiang Li 2 , Pengfei Zhou 1 , Mo Li 1 Nanyang Technological University 1 , Singapore City University of Hong Kong 2 , Hong Kong, China 12 September 2019 UbiComp19,


  1. Amateur: Augmented Reality Based Vehicle Navigation System Chu Cao 1 , Zhenjiang Li 2 , Pengfei Zhou 1 , Mo Li 1 Nanyang Technological University 1 , Singapore City University of Hong Kong 2 , Hong Kong, China 12 September 2019 UbiComp’19, London, UK

  2. � 2 Transportation System Vehicular navigation system Mobile navigation service

  3. � 3 Transportation System Problem: display digital map Gap: real v.s. virtual

  4. � 4 AR-based Navigation Service ✦ Display front-view road condition ✦ Instructions on live world ✦ Comparable navigation ✦ Easy to deploy

  5. � 5 AR-based Navigation Service ❖ Challenges Determine the correct instruction at proper position on screen. ✦ Host-lane identifying ✦ Annotations placement Lane-level localisation accuracy Depth information in video

  6. � 6 AR-based Navigation Service ❖ Related work CarLoc: Precisely Tracking Automobile Real time Detection of Lane Markers in Position [SenSys’15] Urban Streets [IVS’08] 20 built-in sensors Complicated image processing ✦ ✦ Dead-reckoning Too heavy to be affordable ✦ ✦

  7. � 7 AR-based Navigation Service ❖ Related work Tesla autopilot 2.0 Google automobile Maintain on hostlane Rich sensor embedded ✦ ✦

  8. � 8 AR-based Navigation Service ❖ Related work Towards Unified Depth and Semantic Prediction from a Single Image [CVPR’15] Three complicated neural networks ✦ Large volume of training data ✦

  9. � 9 AR-based Navigation Service ❖ System architecture Instructional Sign Placement Lane-Road Information Instruction Type Instruction Position AR Result Host Lane Intersection Navigation Particle Filter GPS Pin-hole Model & Live Motion Origin and Destination Peak Detection Traffic Light Detection video sensor System GPS User Input Frame Slice Subtracter Frames Lane Identification Intersection Inference ce

  10. � 10 AR-based Navigation Service ❖ System architecture Instructional Sign Placement Lane-Road Information Instruction Type Instruction Position AR Result Host Lane Intersection Navigation Particle Filter GPS Pin-hole Model & Live Motion Origin and Destination Peak Detection Traffic Light Detection video sensor GPS User Input Frame Slice Subtracter Frames Lane Identification Intersection Inference

  11. � 11 AR-based Navigation Service ❖ System architecture — lane identification Lane detection task. Lane identification task. Based on pure videos. Based not only on videos… IMU sensors on mobile phone & Avoid collisions for automobiles. Extra lane number information Assistant for drivers.

  12. � 12 AR-based Navigation Service ❖ System architecture — lane identification One frame in video Image slicing of 60 frames

  13. <latexit sha1_base64="SZ0J2yGuG58XxklAN9wP4PnaU=">ACB3icdVDLSsNAFJ34rPUVdSnIYBEsSRtoXZXcKHLCvYBbQiT6aQdOnk4MxFKyM6NfoA/4M6NC0Xc+gvu/AD/w2lSQUPXDicy/3uOEjApGO/azOzc/MJibim/vLK6tq5vbLZEHFMmjhgAe84SBGfdKUVDLSCTlBnsNI2xkdT/z2JeGCBv65HIfE8tDApy7FSCrJ1nd6Lkc4NpO4nISwx8gFjMPDEzs2DTlMElsvGMVaCpiRamVKaiY0i0aKQv3grnX7cVNu2Ppbrx/gyCO+xAwJ0TWNUFox4pJiRpJ8LxIkRHiEBqSrqI8Iqw4/SOBe0rpQzfgqnwJU/X7RIw8Icaeozo9JIfitzcR/K6kXSPrJj6YSJj7NFbsSgDOAkFNinGDJxogzKm6FeIhUsFIFV1ehfD1KfyftEpFs1wsnak0KiBDmyDXbAPTFAFdXAKGqAJMLgC9+ARPGnX2oP2rL1krTPadGYL/ID2+gnkFp0C</latexit> <latexit sha1_base64="9RDv2MYFcOXvS2/hoKTrS5Q6S04=">ACDHicdVDLSgMxFM34rPVdekmWARBKDNtoXYhFNy4rGAf0Kklk2ba0MzDJCOWMB+gC5f+gB/gxoUibv0Ad36A/2FmWkFDwQO5zLzT1OyKiQpvluzMzOzS8sZpayura+u5jc2mCKOSQMHLOBtBwnCqE8akpG2iEnyHMYaTmjo8RvXRAuaOCfynFIuh4a+NSlGEkt9XL5oT0g59B2OcLKilUx9tClrQY9asc9RQ+t+CxJmYVqCjghlfKUVC1oFcwU+dr+XfP247pU7+Xe7H6AI4/4EjMkRMcyQ9lViEuKGYmzdiRIiPAIDUhHUx95RHRVekwMd7XSh27A9fMlTNXvEwp5Qow9Ryc9JIfit5eIf3mdSLoHXUX9MJLEx5NFbsSgDGDSDOxTrBkY0Q5lT/FeIh0r1I3V9Wl/B1KfyfNIsFq1Qonug2ymCDNgGO2APWKACauAY1EDYHAF7sEjeDJujAfj2XiZRGeM6cwW+AHj9ROtdJ+1</latexit> <latexit sha1_base64="Rkf7FqIuQSgMdOn9OZY+TVWMXbI=">ACG3icdZDLSgMxFIYz9VbrerSTVBEQVpmxkLtQi40GUF2yqdWjJpg3NZIYko5Rh3sONj6IbF4q4ElwIfRjTi2BFfwj8fOcTs7vhoxKZqfRmpmdm5+Ib2YWVpeWV3Lrm/UZBAJTKo4YIG4dJEkjHJSVQxchkKgnyXkbrbOxnW6zdESBrwC9UPSdNHU49ipHSqJW1w+NuzqfciTst6iTXce6d+B4AuH4NontJGkNSW6KZHfMfGkODbFwsSULGjlzZF2yvtXp3uDh0GlX132gGOfMIVZkjKhmWGqhkjoShmJMk4kSQhwj3UIQ1tOfKJbMaj2xK4q0kbeoHQjys4oj8nYuRL2fd3ekj1ZW/a0P4V60RKe+oGVMeRopwPF7kRQyqA6Dgm0qCFasrw3Cguq/QtxFOgal48zoEL4vhf+bmp23DvP2uU6jAMZKgy2wDfaBYqgDM5ABVQBnfgETyDF+PeDJejbdxa8qYzGyCKRkfX4oprQ=</latexit> � 13 AR-based Navigation Service ❖ System architecture — lane identification 4 200 5 2 3 6 1 150 Grey value h p 100 7 50 w 0 0 200 400 600 800 1000 1200 1400 1600 1800 Pixel indicator 1 h ≥ 1 g k + w 2 max { g i } n p = h − min { g i } 2 3 p ≤ p − G 10 th g k − w i =1 2 Brightness Sharpness

  14. � 14 AR-based Navigation Service ❖ System architecture — lane identification Lane Marker [790] [1106] [474] Number X 1 [-474] [-474] [474] [474] [790] [790] X 2 Lane [474] [-790] [-474] Direction X 3 [- [-1106] [-790] [-474] X 4 Lane Number A road segment with 4 lanes Templates of a 4-lane road

  15. � 15 AR-based Navigation Service ❖ System architecture — lane identification X obs X 3 [-790] [-474] [474] [474] [-790] [-474] Matching Ideally detect the peaks 1. Blockage of frontal vehicles 2. Reflection of lights 3. Bad condition of lane markers

  16. � 16 AR-based Navigation Service ❖ System architecture — lane identification ✦ Particle filter design [474] [790] [1106] X obs X 1 [-790] [-474] [474] [790] [790] [-474] [474] [-474] [474] [790] X 2 [-790] [-474] [474] Weight updating X 3 [-1106] [-790] [-474] X 4 Dynamic time wrapping initialisation

  17. <latexit sha1_base64="eXOYNe48FvqEsizy0aM4VtsE/i0=">AB9XicdVDLSsNAFJ3UV62vqhvBzWAR3BiSWqhFhIblxXsA9o0TCaTdujkwczEUkL3foIbF4q4danf4c6/cZpUNEDFw7n3Mu9zgRo0IaxoeW1hcWl7JrxbW1jc2t4rbOy0RxhyTJg5ZyDsOEoTRgDQlYx0Ik6Q7zDSdkYXM79Q7igYXAtJxGxfDQIqEcxkrqj2nH52TfnLs2s7ULpYMvZYCZqRamZOaCU3dSFGq791ar6Ozt4ZdfO+5IY59EkjMkBd04iklSAuKWZkWujFgkQIj9CAdBUNkE+ElaRXT+GhUlzohVxVIGqfp9IkC/ExHdUp4/kUPz2ZuJfXjeW3qmV0CKJQlwtsiLGZQhnEUAXcoJlmyiCMKcqlshHiKOsFRBFVQIX5/C/0mrJsnevlKpVEBGfJgHxyAI2CKqiDS9ATYAB3fgATxqY+1e9Kes9acNp/ZBT+gvXwCiCWLQ=</latexit> � 17 AR-based Navigation Service ❖ System architecture — lane identification ✦ Particle filter design Dynamic time wrapping under Euclidean distance X obs X 3 [-790] [-474] [474] [790] [474] [-790] [-474] Matching w p b = e − d b

  18. � 18 AR-based Navigation Service ❖ System architecture — lane identification ✦ Particle filter design Resampling based on importance X obs [474] [790] [-790] [-474] Movement: lane switching

  19. � 19 AR-based Navigation Service ❖ System architecture — lane identification ✦ Particle filter design Resampling based on importance

  20. � 20 AR-based Navigation Service ❖ System architecture — lane identification ✦ Particle filter design Resampling based on importance Lane marker of host-lane has a traversing phenomenon during lane switching.

  21. <latexit sha1_base64="F+5Sx7dI6tf6U3W8FKAEb7P86qI=">ACQHicdVBLSwMxGMz6tr6qHr0Ei9Bel0t1B6EghcPghWsFbqlZNsG5p9kHwrlO3+NC/+BG+evXhQxKsn03YVlToQmMx8Q76MGwmuwLIejbn5hcWl5ZXV3Nr6xuZWfnvnWoWxpKxBQxHKG5coJnjAGsBsJtIMuK7gjXdwenYb94yqXgYXMEwYm2f9ALucUpAS5180/EJ9CkRST0tfvPztKNG37eLtHQyc+wiHf2KlBzaDcFxGZBOvmCZ1QnwlFTKGana2DatCQoQ72Tf3C6IY19FgAVRKmWbUXQTogETgVLc06sWETogPRYS9OA+Ey1k0kBKT7QShd7odQnADxRfyYS4is19F09Od5X/fXG4iyvFYN3E54EMXAjp9yIsFhCP28RdLhkFMdSEUMn1rpj2iSQUdOc5XcLXT/H/5PrQtI/Mw8tyoWZmdaygPbSPishGFVRDZ6iOGoiO/SEXtCrcW8G2/G+3R0zsgyu+gXjI9P65iy9w=</latexit> <latexit sha1_base64="OilGbpP5ivGFm41K08aA2r/hG9I=">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</latexit> <latexit sha1_base64="2WczuQIuT4KbxgclCcSKZ6L9I+0=">ACHXicdVDLSgMxFM3UV62v+ti5CRaxLhxmaqF2IRTcuBCsYGuhHYdMJtOGZh4kGaGM8yNu/BU3LhRx4Ub8G9OHRUPBE7OuZd73EiRoU0jA8tMzM7N7+QXcwtLa+sruXN5oijDkmDRykLcJAijAWlIKhlpRZwg32HkyumfDP2rG8IFDYNLOYiI5aNuQD2KkVSnS93fCR7GLGknhan/DyFt3D6O0tsQ+PIblODlxb7KV2vmDo1RHgmFTKE1I1oakbIxTABHU7/9ZxQxz7JCYISHaphFJK0FcUsxImuvEgkQI91GXtBUNkE+ElYyuS+GuUlzohVy9QMKR+r0jQb4QA9RlcONxW9vKP7ltWPpHVkJDaJYkgCPB3kxgzKEw6igSznBkg0UQZhTtSvEPcQRlirQnArh61L4P2mWdPNQL12UC7WtSRxZsA12QBGYoAJq4BTUQNgcAcewBN41u61R+1Fex2XZrRJzyb4Ae39E8DzojI=</latexit> � 21 AR-based Navigation Service ❖ System architecture — lane identification ✦ Particle filter design Resampling based on importance P ( L s |O ) = P ( O|L s ) · P ( L s ) P ( L s |O ) = P ( O|L s ) · β P ( O ) P ( O|L s ) = e − d 0 s

Recommend


More recommend