News register
Search section
Search area
News Type
News type
Search date
Search word OR
List of related articles
Contents of related articles
No info was found
List of related articles
Contents of related articles
No info was found
View details
Information

09/09/2019 HYU News > Academics > 이달의연구자

Title

[Excellent R&D] Development of Computer Vision Algorithms for Spatial Recognition of Videos

Professor Lim Jong-woo (Department of Computer Science)

김현수

Copy URL / Share SNS

http://www.hanyang.ac.kr/surl/Bg59

Contents
The Next-generation Information Computing Development Project is a research project executed by Hanyang University and six other research teams, which has been ongoing from September of 2017 and will end on December of 2020. There are two main parts of the research, and Professor Lim Jong-woo (Department of Computer Science) took charge of the first part, titled "fundamental study of vision algorithms for spatial recognition of videos." The focus of Lim's research was to develop computer vision algorithms for spatial recognition of videos.
 
Professor Lim Jong-woo (Department of Computer Science) is taking part in the Next-generation Information Computing Development Project.

The object of this research was to develop a computer vision algorithm to comprehensively recognize accurate three-dimensional information of surrounding environments and to detect and predict the location and movement of important figures through the various videos achievable in routine environments. With the basis on geometrical probabilistic computer vision algorithms that have been the subject of research as of now, the research team of six has been striving to develop an original technology that can successively perceive and comprehensively infer information on the environment and major objects inside the video. The first theme consists of geometrical environment information recognition, and the other is detection and tracking of principal objects.
 
Devices with cameras equipped are usually used for taking photos or videos. This research plans to overcome the limitations of the existing methodology, which is the information quantity of the environment map and updating method. They developed a stochastic algorithm that can effectively accumulate long-cumulated information and extract three-dimensional street information of the overall environment by maximizing the information that can be earned from the video. The ultimate goal is to make sure that research output is applied to robots, wearable devices, and autonomous cars by developing an algorithm that accurately model the movements of objects.
 
Original image and restored distance map from blurred image
(Photo courtesy of Lim)

Object detection technology is emerging and is recently being more widely used in research with deep learning to increase the accuracy of detection. To resolve the issue of difficulty in detecting, clearly due to complex interactions between objects, sudden movements or frequent covering of objects, Lim and his research team sought to develop a deep learning based object detecting technology.
 
Lim has looked into geometric vision for about 10 years. He started motion estimations with a camera at Honda research, until in 2011, when he developed a service that enabled the technology to expand to indoors, as part of the street view team of Google. He continued with geometric vision research at Hanyang University from 2012. Now that the first part of research has been completed, Lim revealed that there is still a ways to go, as it is tough for a computer to recognize as well as humans. Nevertheless, aimed functions were developed and published as a thesis. Lim advises people interested in looking into deep learning to learn in a systematic fashion and study carefully. “Deep learning is a strong tool, but it is not almighty.”

 
 
Copy URL / Share SNS

0 Comments