• 정보처리학회논문지:소프트웨어 및 데이터공학
• /
• 제11권12호
• /
• pp.509-516
• /
• 2022

3차원 공간에서 물체들의 정확한 자세 예측은 실내외 환경에서 장면 이해, 로봇의 물체 조작, 자율 주행, 증강 현실 등과 같은 많은 응용 분야들에서 폭넓게 활용되는 중요한 시각 인식 기술이다. 물체들의 자세 예측을 위한 과거 연구들은 대부분 각 인식 대상 물체마다 정확한 3차원 CAD 모델을 요구한다는 한계점이 있었다. 이러한 과거 연구들과는 달리, 본 논문에서는 3차원 CAD 모델이 없어도 RGB 컬러 영상들만 이용해서 미지 물체들의 자세를 예측해낼 수 있는 새로운 신경망 모델을 제안한다. 제안 모델은 적응형 깊이 추정기인 AdaBins를 이용하여 스스로 미지 물체 자세 예측에 필요한 각 물체의 깊이 지도를 효과적으로 추정해낼 수 있다. 벤치마크 데이터 집합들을 이용한 다양한 실험들을 통해, 본 논문에서 제안한 모델의 유용성과 성능을 평가한다.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• 제14권11호
• /
• pp.4557-4572
• /
• 2020

The Alaska competition provides an opportunity to study the practical problems of real-world steganalysis. Participants are required to solve steganalysis involving various embedding schemes, inconsistency JPEG Quality Factor and various processing pipelines. In this paper, we propose a method to ensemble multiple deep learning steganalyzers. We select SRNet and RESDET as our base models. Then we design a three-layers model ensemble network to fuse these base models and output the final prediction. By separating the three colors channels for base model training and feature replacement strategy instead of simply merging features, the performance of the model ensemble is greatly improved. The proposed method won second place in the Alaska 1 competition in the end.

• 한국통신학회논문지
• /
• 제31권11C호
• /
• pp.1071-1076
• /
• 2006

본 논문에서는 수화 인식을 위한 얼굴 및 손 추적시스템을 제안한다. 제안된 시스템은 검출 및 추적 단계로 구분된다. 검출 단계에서는 신호의 주체인 얼굴과 손에 위치한 피부 특징을 이용하였다. CbCr 공간에서의 타원 모델을 구성하여 피부 색상을 검출하고 피부 영역을 분할한다. 그리고 크기와 얼굴 특징을 이용하여 얼굴과 손 영역을 정의한다. 추적 단계에서는 동작 추정을 위하여 첫 번째 손 영역으로 예측된 다음의 손위치를 연산함으로써 두 번째 손의 영역을 유도해낸다. 그러나 갑작스런 움직임의 속도 변화가 있을 경우 연속된 프레임에서 추적된 위치는 부정확하였다. 이러한 점을 해결하고자 손 영역에 대하여 반복적인 재연산을 수행하여 적응적으로 영역을 찾음으로써 오차를 보정하도록 하였다. 실험 결과 제안된 방법은 기존의 방법보다 4%의 처리 시간이 증가된 반면, 예측 오차는 96.87%까지 감소시킬 수 있었다.

• 한국인터넷방송통신학회논문지
• /
• 제23권1호
• /
• pp.65-70
• /
• 2023

본 논문에서는 화재로부터 실시간으로 화염과 연기를 감지하고 분할하기 위해 딥러닝 모델을 사용하였다. 이를 위해 의미론적 분할에서 우수한 성능을 보이는 U-NET을 사용하고 다중 클래스를 이용하여 화재의 불꽃과 연기를 구분 하였다. 제안된 기법을 이용하여 학습한 결과, 손실 오차와 정확도 값이 각각 0.0486과 0.97996으로 매우 양호하였다. 객체 감지에 사용되는 IOU 값도 0.849로 매우 좋았다. 학습된 모델을 이용하여 학습에 사용하지 않은 화재 이미지를 예측한 결과, 화재의 불꽃과 연기가 잘 감지되고 분할되었으며, 연기의 색상도 잘 구분되었다. 제안된 기법을 이용하여 화재 예측 및 감지 시스템 구축 등에 사용될 수 있다.

• 산업경영시스템학회지
• /
• 제45권2호
• /
• pp.48-55
• /
• 2022

The color image of the brand comes first and is an important visual element that leads consumers to the consumption of the product. To express more effectively what the brand wants to convey through design, the printing market is striving to print accurate colors that match the intention. In 'offset printing' mainly used in printing, colors are often printed in CMYK (Cyan, Magenta, Yellow, Key) colors. However, it is possible to print more accurate colors by making ink of the desired color instead of dotting CMYK colors. The resulting ink is called 'spot color' ink. Spot color ink is manufactured by repeating the process of mixing the existing inks. In this repetition of trial and error, the manufacturing cost of ink increases, resulting in economic loss, and environmental pollution is caused by wasted inks. In this study, a deep learning algorithm to predict printed spot colors was designed to solve this problem. The algorithm uses a single DNN (Deep Neural Network) model to predict printed spot colors based on the information of the paper and the proportions of inks to mix. More than 8,000 spot color ink data were used for learning, and all color was quantified by dividing the visible light wavelength range into 31 sections and the reflectance for each section. The proposed algorithm predicted more than 80% of spot color inks as very similar colors. The average value of the calculated difference between the actual color and the predicted color through 'Delta E' provided by CIE is 5.29. It is known that when Delta E is less than 10, it is difficult to distinguish the difference in printed color with the naked eye. The algorithm of this study has a more accurate prediction ability than previous studies, and it can be added flexibly even when new inks are added. This can be usefully used in real industrial sites, and it will reduce the attempts of the operator by checking the color of ink in a virtual environment. This will reduce the manufacturing cost of spot color inks and lead to improved working conditions for workers. In addition, it is expected to contribute to solving the environmental pollution problem by reducing unnecessarily wasted ink.

• 복식
• /
• 제40권
• /
• pp.109-119
• /
• 1998

In order to identify the factors responsible for the recent drastic increase of imported apparel in Korea, an attempt was made to determine the variables influencing the purchasing behavior of imported and domestic apparel and forecast the purchasing intention with the use of Fishbein & Ajzen's behavioral intention model including both attitude and subjective norm. Based on literature review, the empirical study was conducted using the questionnaire for 900 college women and high school girls living in Seoul. Descriptive statistics, t-test, paired-t test, multiple regression analysis, and correlation analysis were made of 771 returned questionnaires using SAS program. The results were as follows : First, the results of assessing both their attitudes toward imported and domestic apparel and their subjective norms were shown to be different. Second, there was a difference in the attributes that had an effect on their attention to buy imported and domestic apparel. Third, those respondents having a preference for imported apparel were most highly influenced by color and price. While those respondents showing a preference for domestic apparel were most highly influenced by materials and comfortableness. Fourth, the validity of the prediction value of their buying intention was confirmed as it was shown to be more than coreelation coefficien r=0.65. In conclusion, 1) it was proved that both attitude and subjective norm were the important variables that could predict the consumer's purchasing intention, 2) since competitiveness in color and materials and brand influencing the consumer's purchase of and preference for domestic apparel relatively lagged behind in comparison with that of imported apparel, the domestic apparel business enterprise will have to make a greater effort to develop differentiated color, material and prestigious brand so as to enhance competitiveness with imported apparel.

• 농업과학연구
• /
• 제48권4호
• /
• pp.891-897
• /
• 2021

There is a need for a technology that can quickly and accurately analyze soil carbon contents. Existing soil carbon analysis methods are cumbersome in terms of professional manpower requirements, time, and cost. It is against this background that the present study leverages the soil physical properties of color and water content levels to develop a model capable of predicting the carbon content of soil sample. To predict the total carbon content of soil, the RGB values, water content of the soil, and lux levels were analyzed and used as statistical data. However, when R, G, and B with high correlations were all included in a multiple regression analysis as independent variables, a high level of multicollinearity was noted and G was thus excluded from the model. The estimates showed that the estimation coefficients for all independent variables were statistically significant at a significance level of 1%. The elastic values of R and B for the soil carbon content, which are of major interest in this study, were -2.90 and 1.47, respectively, showing that a 1% increase in the R value was correlated with a 2.90% decrease in the carbon content, whereas a 1% increase in the B value tallied with a 1.47% increase in the carbon content. Coefficient of determination (R²), root mean square error (RMSE), and mean absolute percentage error (MAPE) methods were used for regression verification, and calibration samples showed higher accuracy than the validation samples in terms of R² and MAPE.

• 한국식품저장유통학회지
• /
• 제24권2호
• /
• pp.181-186
• /
• 2017

본 연구에서는 토마토를 일정한 기간(15일) 동안 저장하여 토마토의 품질인자인 감모율, 색상과 경도를 분석하고 비 파괴적 방법으로 토마토 경도를 측정한 결과로 경도와 감모율 사이의 상관관계를 분석하고 선형회귀모델을 도출하여 토마토 품질을 예측하고자 하였다. 그린하우스에서 재배된 '티와이250'품종을 수확한 후 일반 과실 종이 박스에 포장한 후 설정환경이 $10^{\circ}C$, 90%RH인 항온항습챔버에 저장하여 3일 간격으로 경도, 무게와 색상 변화를 조사하였다. 15일간 저장 중 감모율은 저장기간이 증가함에 따라 증가하나 1.1% 정도에 머물러 단순히 감모율 인자로만 판단했을 때 토마토의 신선도 품질에 영향 끼치는 수준은 아니다. 색상변화 중 명도와 색조 각은 저장기간의 증가에 따라 감소하는 경향을, a/b와 ${\Delta}E$는 증가하는 경향을 나타냈었고 일원분산분석결과를 볼 때 유의한 수준이었다(p<0.001). 경도는 저장기간이 증가함에 따라 감소하는 경향을 나타냈으며 저장 15일차에는 경도감소율이 40% 이상인 것으로 나타났다. 그리고 비 파괴방법으로 측정한 경도 값과 감모율 사이의 상관관계를 분석하여 선형회귀모델 $WL=-0.0241{\times}F+1.5213$을 하였으며 모델의 추정치 오차는 ${\pm}0.231$이었다. 이러한 결과에 비추어볼 때 수확 후 일정한 저장환경에서 비파괴적인 경도 측정을 통해 감모율을 추정하고 신선도를 판단하는 지표고 사용하는 것이 가능한 것으로 판단되었다.

• IEIE Transactions on Smart Processing and Computing
• /
• 제1권2호
• /
• pp.73-77
• /
• 2012

This paper reports a combined depth- and model-based face detection and tracking approach. The proposed algorithm consists of four functional modules; i) color-based candidate region extraction, ii) generation of the depth histogram for handling occlusion, iii) rotation-invariant face region detection using ellipse fitting, and iv) face tracking based on motion prediction. This technique solved the occlusion problem under complicated environment by detecting the face candidate region based on the depth-based histogram and skin colors. The angle of rotation was estimated by the ellipse fitting method in the detected candidate regions. The face region was finally determined by inversely rotating the candidate regions by the estimated angle using Haar-like features that were robustly trained robustly by the frontal face.

• Journal of the Optical Society of Korea
• /
• 제9권3호
• /
• pp.107-110
• /
• 2005

An optical method of measuring hematocrit noninvasively is presented. An LED Light with multiple wavelengths was irradiated on fingernail and transmitted light from the finger was measured to predict hematocrit. A finger probe contained an LED array and detector. Our previous experience showed that prediction accuracy was sensitive to reliability of the finger probe hardware and we optimized the finger probe parameters such as the internal color, detector area and the emission area of a light source based on Design of Experiment. Using the optimized finger probe, we developed a hematocrit monitoring system and tested with 549 persons. For the calibration model with 368 persons, a regression coefficient of 0.74 and a standard deviation of 3.67 and the mean percent error of $8\%$ were obtained. Hematocrits for 181 persons were predicted. We achieved a mean percent error of $8.2\%$ where the regression coefficient was 0.68 and the standard deviation was 3.69.