Research

This page provides a brief introduction on a number of selected research projects that I have done.

Multimedia information processing

eml_example_1.png echronicle.png pct.png mets.png ems.png stereovision.png

  • EML: EML is an E-model data query language extending the SQL-89 standard with enhanced features: (1) unstructured query, (2) semantic query expansion, (3) temporal query, (4) ranked ordering, (5) path query, and (6) natural join. An E-model prototype is implemented on top of relational database. Hundreds of SQL procedures and functions are developed to support full operations. The idea is that a database system adds the E-model database to existing databases, then a user can query the database in the mixture of structured queries and E-model queries. EML implements such ideas with an extended SQL language that supports both structured and unstructured queries with various feature additions.
  • EFIM: EFIM is a prototype information system development project using an event-based function information model. This project pursues to develop a new information engine for information chronicling. To meet complex demands in heterogeneous data archiving, EFIM has a number of unique features: unified data indexing, progressive event logging and trace back supports, behavior pattern generation and matching, and functional summarization in multiple levels of granularity.
  • eChronicle: eChronicle is an information chronicling system that supports general purpose information archiving using its specialized information model and the EFIM back-end. It features using events as its core to organize the information.
  • Personal chronicling tools: PCT supports enterprise knowledge workers for digital event archiving and collaboration-oriented publishing. PCT is composed of four primary tools: (1) event monitoring, (2) interactive annotation, (3) browse/search, and (4) edit/publish. All are designed to exploit existing enterprise infrastructure, storing captured raw data and metadata in secure databases. PCT is developed during my summer IBM research internship in 2004 at IBM T.J. Watson research center.
  • Media event tagging system: METS is a multimedia event processing system that combines human expertise with algorithmic processing capabilities. Our approach pursues to bridging the “signal-to-symbol” barrier which is based on a 3-layered event-processing architecture composed of a lower data event layer, higher domain event layer, and an elemental event layer in the middle, which links data and domains with symbolic indices. METS is supported by HP Labs and I was the main developer.
  • Experiential meeting system: EMS is an experiential meeting system to allow people to be tele-present in a remote meeting and to be able to review proceedings of a meeting or of several meetings using all the data recorded in a meeting. We consider this as a problem in management and experiential access to all multimedia data acquired in a meeting.
  • Hierarchical mean field algorithms for stereo vision: HMEAN builds the 3-D disparity map from two binocular images. It is featured in three contributions: (1) horizontal and vertical line processes to detect the matching point, (2) hierarchical architecture to minimize the disparity errors in MRF(Markov random fields) modeled 3-D worlds, (3) robust parameter optimization techniques.

GIS information processing

traffic_sign_coder_0824_1.png

  • Container ship pattern recognition: The CSPR project is to recognize the container ship entering the harbor gate to control their access and manage according harbor assets to have the traffic flow efficient and cost-minimizing purposes. Our team is currently working with the U.S. States government engineering team to exploit its usage on the the Georgia Savannah harbor. This project is launched in 2007 and currently under extensive research reviews to specify its topics and coverage.

Software development (selected)

rmsim.png

  • Roughing mill process design software (RMSIM): RMSIM is an enginerring software package composed of 0.25 million lines MFC/C++ codes with hundreds of classes. It gives users the power to design the complex roughing mill plant design. Its simple interface with highly automated parameter setting features astonished users in the industry. At the time it was developed, RMSIM was the only software in the world that supports the full roughing mill plant design, plant simulation and computing optimal production schedules. This is funded by POSCO for 3 years (1999-2001) with $0.5 million and I, as the main software developer in this project, converted existing FORTRAN simulation codes into C++ and all other GUI interfaces.
  • Finishing mill setup simulator (FSUSIM): FSUSIM is a finishing mill line simulation software which is built on top of real plant operation data. FSUSIM provides fully interactive GUIs for users as if they actually set up the product schedule with the real operation data. This software was at first developed for the research simulation and later being populated by the workers at the plant. I was invited to give talks and educate users on FSUSIM three times. FSUSIM is funded by POSCO for two years (1995-1996) with $0.2 million budget.
  • Heavy ingot reduction scheduler (HIRS): HIRS is an engineering software producing the optimized ingot reduction schedule. HIRS algorithms are developed by extensive FEM (Finite Element Method) simulation results with two years on-line operation data.
  • Scale prediction software for finishing mill and downcoiler (SPFM): SPFM predicts the scale thickness during air-cooling time when a downcoil passed the final finishing mill line in the hot-coil and is placed at the yard for cooling. SPFM simulates 3-dimensional temparature environments using non-linear computation libraries. We developed the main SPFM algorithm with extensive engineering data for simulation. SPFM comes with 3D analysis and display packages to visualize the growth of scales on coil over time.

Industrial automation (selected)

warp.png

  • Real-time hot slab warping control system (RTWARP): RTWARP is a real-time video processing system working in hazardous steel plant environments. RTWARP also controls the roughing mill system to flatten the warp that causes serious impair of the roll and even stopping the steel production. RTWARP detects and measures the curvature of the slab end in real time and controls the milling speed in real time. The RTWARP project was funded by POSCO for 3 years with $1.5 million budget. I joined RTWARP at its 2nd year (1994) as the main hardware and software engineer to developed the next generation RTWARP to enhances its performance and stabilize the system. RTWARP won several major awards and also copied to other steel plants successfully.
  • Slab size meter using laser range sensors (SSM): SSM is a real-time sensing system prototype to measure the size of a moving slab in real time using three laser range sensors and one tachometer. SSM was built in 1995 at POSCO and achieved 0.1 mm accuracy in range measurement.
  • Precise roll gap setup technology for edger roll in roughing mill (PRGS): PRGS is developed to measure the diameter of the edger roll in roughing mill to set up the slab width with very high accuracy. It is installed directly under the rough mill and regularly measures the edger gap width. The roll gap is measured using the robotic arm equipped with LVDT (Linear Vertical Displacement Transducer) and built on top of the hydro-powered telescopic cylinder for up and down operation.