Transcript Document
Capstone: Vision, Video and VR CSC 59866CD Fall 2004 Lecture 2 Project Topic 1: Augmented NYC Zhigang Zhu, NAC 8/203A http://www-cs.engr.ccny.cuny.edu/~zhu/ Capstone2004/Capstone_Sequence2004.html Augmented NYC: Interaction, Modeling and Visualization of a Human-In-the-Loop Traffic Sensor Network Or CityEyes: City Traffic Sensor Network for Everyone and by Everyone for a Safer Life Zhigang Zhu Department of Computer Science The City College / CUNY 7/16/2015 Copyright @Zhigang Zhu 2 Outline A Bigger Picture The Real-World Problem Capstone Project 7/16/2015 Scale-down project Visualization and Vehicle Extraction parts Copyright @Zhigang Zhu 3 A BIGGER Picture 7/16/2015 Copyright @Zhigang Zhu 4 Related Research Efforts at CUNY Prof. Prof. Prof. Prof. Prof. 7/16/2015 Robert Haralick, PR Lab Brian Schwartz, New Media Lab Robert Paaswell, UTRC II George Wolberg, Vision & Graphics Zhu – Video Computing Copyright @Zhigang Zhu 5 Background NYC DOT’s Traffic Management Center DOT’s Real-Time Traffic Cameras 86 closed circuit TV cameras on major arteries of NYC 44 video streams or updated still images on the web City Drive Live on Crosswalks Channel 74 7/16/2015 22 live streams, weekday mornings from 6:30 to 9:00 am Copyright @Zhigang Zhu 6 NYC DOT’s Traffic Management Center 86 cameras NOW more in the future 11,000 signals in NYC ITS 7/16/2015 Copyright @Zhigang Zhu 7 Real Player plug-in At times when demand is high, video streaming may not be available to each visitor Live camera locations will be limited during the introductory phase of this new feature Still web cam photos are available at all locations. 7/16/2015 Copyright @Zhigang Zhu 8 City Drive Live on Crosswalks Channel 74 Twenty-two of DOT's traffic cameras Channel 74 on weekday mornings from 6:30 to 9 AM separated views cannot give global information raw video streams Problems (Internet/TV): bandwidth camera failures unavoidable shaking NO (little) processing/ extraction Solutions New ways for Information extraction/compression enhancement integration and presentation 7/16/2015 Storm Feb 18 2003 Copyright @Zhigang Zhu 9 Sensor Network for NYC Traffic Sensor network in 4D (3D space + Time) Traffic Cameras (above ground) Inductive Loops (underground) Inspection Cameras (under vehicles) Aerial Cameras (high in air) Sensors (GPS, EZ-Pass) on Vehicles (mobile) 7/16/2015 Copyright @Zhigang Zhu 10 Research Issues – in cope with NSF Sensors and Sensor Networks localization versus distributed versus centralized processing information extraction on-sensor self-calibration Adaptive management and control of sensor nodes human-in-the-loop or close-loop adaptive feedback 7/16/2015 Copyright @Zhigang Zhu 11 Human In the Loop DARPA Total Information Awareness Google Search Engine Relevance feedback Amozon.com Humans in all the three stages – sensors, alarm and response recommends books based upon what other customers have selected Can we do it in ITS? 7/16/2015 Copyright @Zhigang Zhu 12 Human Involvement in ITS How to get user feedbacks? Web browsers Drivers 7/16/2015 As of Feb 18, 2003, the hit of the San Diego Area Traffic Page of Caltrans since 1996 is 385,379,722, which means more than 100 visits every minute! Real traffic information from sites via wireless connection Copyright @Zhigang Zhu 13 Information Sharing Homeland security Department “ would make this a top priority ” Lockheed Martin Transportation Systems 7/16/2015 one of the biggest challenges confronting ITS today is information sharing Data and control Copyright @Zhigang Zhu 14 How to make these happen? Advanced HCI Interaction ( by everyone) Visualization (for everyone) 7/16/2015 Not only provide info to, but also collect info from people User Internet !! 3D immersive and augmented interface Instead of locator map + video streams Copyright @Zhigang Zhu 15 Proposed Research Traffic Information Extraction (Front End) Traffic Information Analysis (in Background) Vision-based vehicle detection and tracking Human “sensors” - feedbacks For traffic management and advisory Distributed inference model (DBNs) Traffic Visualization (Human Interface) Augmented and immersive user interface 7/16/2015 3D model + real video 2D Map + animation Copyright @Zhigang Zhu 16 Outcome of the Research Traffic Control and Management (DOT) signal timing, city evacuation,... City and Traffic Show (Web, TV) realtime, interactive, 3D walkthrough Traffic Forecasting/Advisory (Web, TV) predicting the future traffic, personalized advisory Traffic Open Source for Education and Research (Internet) interdisciplinary – sensors, vision, graphics, mm, transportation, networking, virtual reality, decision making system... open for students and researchers! 7/16/2015 Copyright @Zhigang Zhu 17 Camera views registered with map…and take a virtual walk.. 7/16/2015 Copyright @Zhigang Zhu 18 Grid Traffic Project Brian Schwartz and Lei Zhou Gridlock simulation Cellular automaton 7/16/2015 Copyright @Zhigang Zhu 19 Virtual Fly-through + Real Video Zhu & Schwartz (CUNY) 2003 7/16/2015 Copyright @Zhigang Zhu 20 Traffic Information Extraction Z. Zhu, et al, VISATRAM: A real-time vision system for automatic traffic monitoring, Image and Vision Computing Journal, 18(10), July 2000: pp 781-794 7/16/2015 Copyright @Zhigang Zhu 21 Security – Under Vehicle Inspection Full FOV coverage when a car drives over an a array of cameras ... One of the cameras P. Dickson, J. Li, Z. Zhu, A. Hanson, E. Riseman, H. Sabrin, H. Schultz and G. Whitten, Mosaic generation for under-vehicle inspection. IEEE Workshop on Applications of Computer Vision, Orlando, Florida, Dec 3-4, 2002, p 251- 256. 7/16/2015 Copyright @Zhigang Zhu 22 Video Mosaics and Bird-Eye View Z. Zhu, E. M. Riseman, A. R. Hanson, Parallel-perspective stereo mosaics, the Eighth IEEE International Conference on Computer Vision, Vancouver, Canada, July 9-12, 2001, vol I, 345-352. 7/16/2015 Copyright @Zhigang Zhu 23 Dynamic Mosaic from 2 Cams Wide FOV movie 7/16/2015 Copyright @Zhigang Zhu 24 Sensors on Vehicles Video Stabilization and GeoRegistration J. S. Jin, Z. Zhu, G. Xu, A stable vision system for moving vehicles, IEEE Transactions on Intelligent Transportation Systems, vol. 1, no. 1, March 2000, pp 32-39 7/16/2015 Copyright @Zhigang Zhu 25 Vehicle Extraction [YeMing Ye, Robert M. Haralick and Linda G. Shapiro, " Estimating Optical flow Using a Global Matching Formulation and Graduated Optimization ", I6th International Conference on Image Processing, 2002 7/16/2015 Copyright @Zhigang Zhu 26 System Diagram 7/16/2015 Fig.6. System diagram of the human-in-loop distributed computing Copyright @Zhigang Zhu 27 Traffic Information Extraction from 4D distributed environment (PAST) spatially distributed sensors over time provide real-time information (NOW) predict the future traffic (FUTURE) Challenging Research Issues 7/16/2015 various road situations, realtime optical flow, traffic volume/velocity, 24/7, 4 season, rain &dry Copyright @Zhigang Zhu 28 Traffic Information Analysis Problems Traffic Information Gridlock, traffic jams ... NOT ONLY space-time data from sensors BUT events, people in NYC! Research Issues 7/16/2015 Mathematic model – differential equations or Simulation – cellular automata Copyright @Zhigang Zhu 29 City and Traffic Show Three modes 2D map + cartoon-like animation (w/ info) 3D model + real video 3D model + (real video) + traffic info Research Issues 7/16/2015 camera calibration...static and dynamic real-virtual registration... geo-referenced fast rendering ... realistic?? new media presentation – enhanced video?? Copyright @Zhigang Zhu 30 Traffic Forecasting (Advisory) Two modes On TV (like weather report) Internet (interactive) give the best route WHEN start ... end WHERE start ... end Research Issues show the animation prediction data from sensors data from models (events, facilities) presentation pseudo-colored real video on map interactive and personalized 7/16/2015 Copyright @Zhigang Zhu 31 Traffic Education Open Source interdisciplinary ubiquitous sensors in NYC vision, graphics, mm ,virtual reality networking, video transmission transportation, traffic, city, tourism control, decision making system... open for students and researchers! study various traffic problems 7/16/2015 Copyright @Zhigang Zhu 32 Related Work Blending simulations with data from sensors Schreckenberg in Germany for up-to-the minute traffic show and 1-hour traffic forecast (http://traf2.uniduisburg.de/OLSIM) Texas A&M for computerized traffic control Source : 1999 Science News Online 7/16/2015 Copyright @Zhigang Zhu 33 Come Down to the Capstone… 7/16/2015 Copyright @Zhigang Zhu 34 What you need to do Augmented NYC demo code: starting point CCNY\Working\NYC_VC\AugmentedNYC\AugmentedNYC.exe C++ and OpenGL NeHe OpenGL Tutorials (with code samples) http://nehe.gamedev.net/ To-Do List – Real Work of the Senior Project 2D or 3D NYC map and virtual walk-through (basics) Multiple Real Video Streams in 3D (basics) Vehicle Extraction (advanced) Traffic Analysis and Prediction (optional) Reports and Papers 7/16/2015 Copyright @Zhigang Zhu 35 Next… Project Topic 1 Update Everyone contributes to the update of the system design Project Topic 2 7/16/2015 Virtualized Classroom Copyright @Zhigang Zhu 36