Sangjoon Lee

A Development and Clinical Evaluation of Automated Diagnostic Algorithm for Atrial Fibrillation using 12-lead EKG

The atrial fibrillation (AF or A Fib) has been known as an important risk factor for stroke. The mortality of patients with AF is known to increase two times compared with the patients with no AF. Moreover, the association of heart failure with AF increases the rates of mortality and hospitalization very high.

Development of an automatic sorting system for fresh ginsengs by image processing techniques

This study was conducted with the objective of implementing a smart IoT (internet of things) factory consisting of an automatic 6-year-old fresh ginseng grade classification device. Conventionally, washed 6-year-old ginseng from farmlands is manually sorted into three grades using classification criteria such as weight and shape. However, the cost associated with this classification process has been on the increase. Consequently, to reduce this associated cost, we developed an automatic ginseng sorting device that classifies 6-year-old ginseng according to weight and shape via image processing and sends the classification results to a factory server over a network. Evaluations conducted of the performance of the developed machine using 100 units of 6-year-old ginseng showed that it has a high recognition rate, with an accuracy of 94% for Grade 1, 98% for Grade 2, and 90% for Grade 3.

Development of vision system for defect inspection of electric parts in the tape and reel package

Tape and reel packaging process is one of the important final manufacturing process of electric chip. During and after this packaging process, the defects of chips in the tape and reel package should be inspected. It is so time and cost consuming task by human inspectors. The main defects of chip in the tape and reel package are orientation chip, a wrong type chip and marking defects of chips. In this paper, the automatic vision system and image processing algorithms are developed to inspect the above defects on tape and reel packaging process. The high speed vision algorithm is based on multi-thread and parallel processing method to meet the inspection time(minimum 200 chips/min). Experiments were performed to evaluate the proposed vision inspection algorithms. The experimental result shows that our algorithm achieves high accuracy and speed for industrial requirement.

Implementation of an Image Restoration with Block Iteration Method for Spatially Variant Blur Models

In image restoration, for spatially variant blur, block-wise iterative method have been proposed. Block-wise iterative method is based on the assumption that the blur approximates to spatially invariant in a small region of the blurred images. These approaches show approximation errors and block artifacts. In this work, we suggest to use block iterative method without approximates to spatially invariant blur considering arbitrary shaped blocks related to shapes of blur models. We can reduce the approximation errors and block artifacts with high speed of iteration convergence. This proposed image restoration method can reduce the computational efforts and is useful to embedded system such as mobile phone and embedded vision system.

Design of force measurement module for force touch screens

This paper presents a foce touch technology that uses strain gauge sensors for next-generation 3D touch screens. We developed a force measurement module using strain gauges, and configured a system that was identifiable according to touch pressure, and tested the force measurement module. We found that the thickness of the base pad in the force measurement module is very important for touch sensitivity, and carried out thickness optimization of the base pad. We successfully conducted an experiment to display touch signals on a screen with various touch pressure and multi-touch testing. This study confirmed the possibility of applying to the force touch screen using the force measurement module.

Segmentation and Counting of Cell in Fluorescence Microscopy Images Using Improved Chain Code Algorithm

This study aims to automatically segment of oval cell in fluorescence stained cell image and quantify cell counts. For this study, an algorithm for oval cell contour tracking was suggested based on the classic chain code method and overlapped cells were segmented using border line angle variation information. For verifying the accuracy of the suggested method, our method and Freeman’s chain code method were applied to the same oval cell images. Then the border line tracking results were identified and the execution speed and computation per pixel were compared. Also, it was compared with the segmentation result of the Watershed technique, which is a general region-based segmentation, for evaluating the cell segmentation result with the naked eye. We applied an automatic algorithm to quantify cell counts in 20 cell images. For verifying the accuracy of cell counting, our algorithm was compared with the result of the manual counting method and ImageJ tool-based counting method.

Development of the Vision System and Inspection Algorithms for Surface Defect on the Injection Molding Case

The surface defects of plastic after the injection molding process have been inspected human inspector. In this paper, we propose the automated surface inspection system to detect 3D defects on injection molding objects. This vision system is equipped with 4-cameras and 8-illuminations and is designed to find optimal configuration between camera, illumination and object to detect the defects on an object. Also we develop algorithms to detect defects in captured images. The developed surface inspection system reduce the labor costs and time, whereas increase the detection accuracy of the surface defects such as scratch. The proposed algorithm uses the difference of the brightness of the obtained image by the camera and illumination system for acquiring data of the surface of the object. We focus on the surface defect detection of black molding object. Scratch detection is performed by the developed vision inspection system.

Implementation of a Smart IoT Factory Using an Agricultural Grade Sorting Device

The purpose of this study is to implement the smart IoT factory through developing an automatic 6-years fresh ginseng grade classification device. The washed 6-years ginseng from farmland should be sorted 3 rating by a classification criterion such as weight and shape but this classification process has been conventionally performed manually and there are increasing sorting process costs. To overcome this disadvantage, we developed an automatic gin-seng sorting device. The 6 years ginseng put into the device for the classification is designed to perform such as the weight estimation and the shape analyzed by the image processing procedure, and classification result is sent to the factory server over the network. Evaluating the performance of developing machine experiment with 100 6-years ginseng showed a high recognition rate for 94 % for grade 1, 98 % for grade 2, and 90 % for grade 3.

2D/3D Visual Optical Inspection System for Quad Chip

In the manufacturing process of the LQFP/TQFP (Low-profile Quad Flat Package/Thin Quad Flat Package), the requirement of a 3 dimensional inspection is increasing rapidly and a 3D inspection of the shape of a chip has become an important report of quality control. This study developed a 3 dimensional measurement system based on PMP (Phase Measuring Profilometry) for an inspection of the LQFP/TQFP chip and image processing algorithms. The defects of the LQFP/TQFP chip were classified according to the dimensions. The 2 dimensional optical system was designed by the dorm illumination to achieve constant light distribution, In the 3 dimensional optical system, PZT was used for moving 90 degree in phase. The problem of 2 ambiguity was solved from the measured moir? pattern using the ambiguity elimination algorithm that finds the point of ambiguity and refines the phase value. The proposed 3D measurement system was evaluated experimentally.

Development of an eye tracking medical headlamp.

This study investigates a medical headlamp which is designed to adjust LED light automatically on the X axis and Y axis plane as determined a surgeon during a surgery. The needs of surgeons are determined through a CMOS camera which detects the movement of the eye. The system is composed of a mechanical actuator and an electrical control center. The LED lamp actuator (50×24×21 mm3) was designed to be as small as possible. The actuator can be easily attached onto typical goggles and incurs no burden during long-term use due to its light weight of 40g. To determine the movement of the eye, a limited embedded CPU was designed while considering memory, the process speed, simplicity and accuracy. The designed system was tested by exercises such as moving the eye in a zigzag direction and in random up, down, left, and right directions with adjustments for different lightings. Each was tested 100 times (total 1500), showing satisfactory results with 98.72% accuracy.