This is a recording of Global CENTRA Webinar on 3 March 2018 (at 8-9pm EST). The webinar series is hosted by CENTRA Project (globalcentra.org, supported by US NSF ACI Award 1550126), headquartered at the **ACIS Lab, University of Florida.
Topic: Camera Sensor Technology and Automatic Traffic Control
Speaker: Dr. Hyuk-Jae Lee, Seoul National University, South Korea
00:00 Presentation begins
00:49 Outline of this talk
00:56 "Autonomous Driving Shuttle Bus" project
02:38 Video monitoring - traffic control with surveillance cameras; camera processing network; moving object detection; localization; simulation platform at the lab
08:50 Video Analysis Service and System - Citilog
11:10 Video Analysis Service and System - Dahua Technology; iteris; Gridsmart
12:06 Ko-PER Research Project
14:12 Video Analysis Operations - from image processing to decision level; video compression, machine learning based vehicle detection and vehicle characteristic recognition
17:44 Vehicle Behavior Analysis: Example
19:12 Latency Critical for Traffic Control
22:22 Traffic Control on an Edge Device
23:41 Edge Device Hardware Structure
25:00 Latency Analysis in an Edge Device - processing sequence
26:30 Latency examples: Object Detection, Vehicle Type Classification, and Digital Image Stabilization
32:28 Latency Increase by DIS; FPGA (Field Programmable Gate Array) to avoid DRAM access latency
35:10 Microsfot Project Brainwave
35:45 Multiple Camera Operations: multiple camera fusion and sensors; object detection; image synchronization
43:15 Communication between Server/Edge
43:52 Summary
46:29 Q & A begins
55:38 Taking the virtual group photo
Abstract:
Automatic traffic control requires various types of sensors including GPS, LIDAR, RADAR, and camera sensors. Theses sensors are used together to detect, track and locate objects of interest such as vehicles, pedestrians, and traffic signs. A camera plays a main role in traffic control because it offers a cost-effective solution for object detection and tracking. This talk introduces the operation principle of a camera sensor capturing an image, enhancing its quality, and compressing the data for transmission and storage. The characteristics for automotive applications are discussed and then recent advances in this research field including the use of AI techniques are presented. Object detection and tracking with cameras are introduced together with the localization of the detected object. This talk also presents the experiences to use cameras for automatic traffic control for self-driving vehicles and discusses the effectiveness of traffic control with cameras.
**Advanced Computing and Information Systems Laboratory (ACIS Lab): acis.ufl.edu/
Follow us on Facebook: facebook.com/GlobalCENTRA and facebook.com/acis.lab
Twitter @GlobalCENTRA