PROJECTS

Research Contents

Research Contents

PROJECTS > Research Contents

2-3 Development of event data recording device and its platform

Participants
TOPIC 1 : Development of Open Platform Event Logger (OPEL) software-based vehicle device and its platform
Participants
  • Jinyoung Yang, CISS
  • Kang Yi, Handong Global Univ. (Computer Science and Electrical Engineering)
  • Dong Sam Ha, Virginia Tech.
Purpose
  • Development of low-power and reliable vehicle blackbox system based on OPEL software platform
  • Development of devices that collect vehicle operation data and sensor data in the vehicle
  • Development of ADAS algorithm using front and rear camera, and product application
Research Contents
  • Development of blackbox device for vehicle
    - FullHD 2CH@2.3W based low price and high performance AP with secured technologies
    - Device for collecting vehicle oepration data and in-vehicle environment data
    - In-vehicle infortainment device based on OPEL platform
  • Development of ADAS algorithm using front and rear camera, and product application
    - Development of robust BSD (Blind Spot Detection) algorithm using a rear camera
    - Development of dangerous situation detection system using a rear camera
    - Development of LDW (Lane Departure Warning), FCW (Forward Collision Warning), FVSA+ (Front Vehicle Start Alarm) algorithm using front and rear camera
  • Product application of battery SoH estimator

Expected Contribution
  • Create services through core technologies of related project’s result
  • Increase market size of vehicle blackbox
  • Secure connecting tool with cloud service using a smartphone
TOPIC 2 : Development of OPEL based personal and stationary device and its platform
Participants
  • JW Choi, CISS
  • Seongchul Lee, CISS
  • Yunju Baek, Pusan National Univ. (Computer Science and Engineering)
Purpose
  • Development and commercialization of smart event-detecting stationary camera and mobile camera system based upon deep learning framework
  • Development of the wearable medical device to prevent sudden infant dead when infants roll over
  • Development of the smart personal camera technology and prototyping
    - Developing the smart personal camera device technology for life logging and safe monitoring
    - A smart personal camera is a platform for extracting the event information by using camera with sensors and passing video with event data using wireless communication
    - Smart personal camera includes the following features:
      ㆍReal-time monitoring of the surrounding environment by the camera and store-and-forwarding for related
        data
      ㆍThe ability to share the collected event information to the cloud server and smart device using wireless
        communication
      ㆍTracking the location of the device using sensor or wireless network
    - Open platform based event logger device
Research Contents
  • Development of stationary camera and mobile camera system
    - Development of AI(artificial intelligence) engine for detecting various events from camera images
    - AI engine optimization to be operable in the resource constrained embedded system
    - Development of mobile camera system with low power mobile processor with cloud service

  • Development of the wearable healthcare system
    - Development of flexible interface for measuring multi-vital informations of infant
    - Development of sensor node for high digital signal processing (The standard of performance evaluation for
      medical device of home healthcare)
    - Development of real-time algorithm for removing motion artifact and sampling vital signal informations
    - Development of wireless device for healthcare standard
    - Development of PC simulation tool and android application for wearable device
    - Design of comfort and portable sock and wearable device wearing on infant ‘foot’

  • Development of smart personal camera platform
    - Development of portable hardware platform using small low-end processor
    - Analyzing the camera images through computer vision library and extracting the events
    - Communicating with smart devices and cloud server via a Wi-Fi based wireless network
  • Development of event logging and transmission technology
    - Extracting the custom event information using video camera and various sensors
    - Establishing action plan by setting desired event
    - Storing the extracted event information to the inside and passing the data over the wireless network
  • Development of cloud-based service
    - Storing the extracted event information in the cloud server
    - Querying the events and video using user's smart device or computer
    - Providing cloud service considering the mobility of the portable devices
  • Development of the software technology
    - Designing the service platform, user interface and related technology

Expected Contribution
  • Contribution on forming and expanding AI-based intelligent CCTV eco-system by providing and developing AI-based event systems and engines
  • U-Health application & accelerated control tech. based on neural signals
  • Continuous demand creation through preoccupancy of healthcare market
  • Easy to provide camera-based monitoring and management service
    - Offering a variety of conveniences by day-to-day recording and tracking of individuals requiring surveillance service
    - Preventing theft of personal belongings and reducing financial losses
TOPIC 3 : Development of OPEL software platform including simulator and Kernel-level OPEL optimization
Participants
  • Dongkun Shin, Sungkyunkwan Univ. (Information and Communication Engineering)
  • Jongmoo Choi, Dankook Univ. (Computer Science and Electrical Engineering)
Purpose
  • Optimization of OPEL Platform, Camera, Sensor, and P2P Communication
  • Robust file system and SD card lifetime estimation technique for OPEL
Research Contents
  • Performance optimization of OPEL(Open Platform Event Logger)’s JavaScript application platform, image processing framework, sensor framework, and P2P connection



  • Developing FAT-compatible robust file system. Partitioning/Pre-allocation for performance enhancement. Checksum-based file-level integrity enforcement
  • Developing SD card lifetime estimation technique. Based on write traffic and read latency, No HW modification. WAF(Write Amplification Factor) consideration. Smartphone App for User interface.

Expected Contribution
  • Introducing the low-power and high-performance IoT service platform based on sensors, cameras, and P2P communication
  • Implementation dependable file system for OPEL devices (e.g. Blackbox,IoT). Monitoring Flash memory based Storage lifetime and health condition
TOPIC 4 : Development of Deep-Learning technologies
Participants
  • Jinwoo Shin / Han, Dongsu, KAIST (Electrical Engineering)
  • Tae-Seong Kim / Sang Min Lee, Kyung Hee Univ. (Biomedical Engineering)
Purpose
  • Development of core technology for detecting anomalous events through images/videos with high accuracy
  • Development of core technology for processing large-scale data in real-time
  • Develop IMU-sensor based wearable device capable of human activity recognition and exercise information generation
  • Develop deep learning based human activity recognition techniques
  • Develop exercise information generation techniques from recognized human activities
  • Develop a wearable device combined with IMU sensors, signal processing module, and wireless communication modules
  • Develop a flexible and stretchable platform for a small patch-type wearable device
  • Develop a wearable device capable of life logging
Research Contents
  • Implementing and training an event classification model using deep CNN architectures(Alexnet or GoogLeNet)
  • Extending to multi-framed input model considering properties of video data
    - The fusion of frame information changing over time: implementation of single frame, early fusion, late fusion, and gradual fusion



    - Multi-resolution CNNs: Boosting up the training and classification while maintaining its performance
  • Data augmentation for improving the performance of classification and collecting additional data
    - Large scale image collection from the Web
    - Improvement of classification performance using noisy image data
  • Applying localization used in ImageNet competition
    - Train a deep network for event classification, and fine-tune the network by replacing softmax output layer with bounding box regressor.
    - Utilize feature maps and weight parameters of trained classification deep network for localization.



  • Develop deep leaning based human activity recognition technology: Convolutional Neural Network (CNN) and Recurrent Network (RNN) based human activity recognition techniques
  • Develop human exercise information generation technology: energy expenditure, step counts, exercise distance etc.
  • Develop software libraries and SDK based on C/Python for human activity recognition
  • Develop IMU sensor based wearable devices capable of human activity recognition and exercise information generation
  • Develop flexible and stretchable platforms for IMU sensor based wearable devices



Expected contributions
  • Source technology development for real-time video detection
  • Leading academic-industrial cooperation through technologies applicable to various areas such as crackdown on illegal parking, antitheft, etc.
  • Smart sensor based human activity recognition technology can contribute automated life logging services
  • Life logging allows personalized life care and health care services
  • Patch-type flexible and stretchable wearable devices can open new opportunities of human activity recognition and life logging services
  • Smart wearable devices and life logging technologies can create new business markets
맨위로

Warning: Unknown: Your script possibly relies on a session side-effect which existed until PHP 4.2.3. Please be advised that the session extension does not consider global variables as a source of data, unless register_globals is enabled. You can disable this functionality and this warning by setting session.bug_compat_42 or session.bug_compat_warn to off, respectively in Unknown on line 0