• 한국산업융합학회 논문집
• /
• 제26권1호
• /
• pp.113-119
• /
• 2023

In this paper, we propose an optimization of a pose estimation deep learning model for recognition of abnormal behavior in unmanned stores using radio frequencies. The radio frequency use millimeter wave in the 30 GHz to 300 GHz band. Due to the short wavelength and strong straightness, it is a frequency with less grayness and less interference due to radio absorption on the object. A millimeter wave radar is used to solve the problem of personal information infringement that may occur in conventional CCTV image-based pose estimation. Deep learning-based pose estimation models generally use convolution neural networks. The convolution neural network is a combination of convolution layers and pooling layers of different types, and there are many cases of convolution filter size, number, and convolution operations, and more cases of combining components. Therefore, it is difficult to find the structure and components of the optimal posture estimation model for input data. Compared with conventional millimeter wave-based posture estimation studies, it is possible to explore the structure and components of the optimal posture estimation model for input data using genetic algorithms, and the performance of optimizing the proposed posture estimation model is excellent. Data are collected for actual unmanned stores, and point cloud data and three-dimensional keypoint information of Kinect Azure are collected using millimeter wave radar for collapse and property damage occurring in unmanned stores. As a result of the experiment, it was confirmed that the error was moored compared to the conventional posture estimation model.

• 사물인터넷융복합논문지
• /
• 제9권2호
• /
• pp.71-76
• /
• 2023

본 논문에서는 단일 수신기가 사용되는 환경에서 정확한 거리 추정을 위해 심층 인공신경망 (Deep Neural Network, DNN)을 활용한 Smoothed Received Signal Strength Indicator (RSSI) 기반 거리 추정 기법을 제안한다. 제안 기법은 거리 추정 정확도 향상을 위해 Data Splitting, 결측치 대치, Smoothing 단계로 구성된 전처리 과정을 수행하여 Smoothed RSSI 값을 도출한다. 도출된 다수의 Smoothed RSSI 값은 Multi-Input Single-Output(MISO) DNN 모델의 Input Data로 사용되며 Input Layer와 Hidden Layer를 통과하여 최종적으로 Output Layer에서 추정 거리로 반환된다. 제안 기법의 우수성을 입증하기 위해 제안 기법과 선형회귀 기반 거리 추정 기법의 성능을 비교하였다. 실험 결과, 제안 기법이 선형회귀 기반 거리 추정 기법 대비 29.09% 더 높은 거리 추정 정확도를 보였다.

• 예술인문사회 융합 멀티미디어 논문지
• /
• 제7권3호
• /
• pp.339-349
• /
• 2017

본 논문에서는 주행 중인 차량의 차선 인식을 위해 4단계로 구성된 알고리즘을 제안한다. 첫 번째 단계에서는 관심영역 추출한다. 두 번째 단계에서는 신호 잡음을 제기하기 위해 중간 값 필터를 이용한다. 세 번째 단계에서는 입력되는 이미지의 배경과 전경의 두 클래스로 구분하기 위한 이진화 알고리즘을 수행한다. 마지막 단계에서는 이진화 과정 후에 남아 있는 노이즈나 불완전한 에지 등을 제거하여 선명한 차선을 얻기 위해 이미지 침식 알고리즘을 이용한다. 하지만 이러한 차선 인식 앍고리즘은 높은 계산량을 요구하여 실시간 처리가 어려운 실정이다. 따라서 본 논문에서는 멀티코어 아키텍처를 이용하여 실시간 차선이탈 감지 알고리즘을 병렬구현 한다. 또한, 차선이탈 감지 알고리즘을 위한 최적의 멀티코어 아키텍처의 구조를 탐색하기 위해 총 8가지의 서로 다른 프로세싱 엘리먼트 구조를 이용하여 실험하였고, 모의실험 결과 40×40의 프로세싱 엘리먼트 구조에서 최적의 성능, 에너지 효율 및 면적 효율을 보였다.

• 문화기술의 융합
• /
• 제10권3호
• /
• pp.713-722
• /
• 2024

본 연구는 인공지능(AI) 기술의 발전이 저널리즘 분야에 가져온 혁신과 변화를 조명하고, 이로 인해 발생하는 주요 윤리적 쟁점들을 검토하여, 저널리즘 분야에서 AI의 활용성을 논의한다. 블룸버그, 가디언, 월스트리트저널(WSJ), 워싱턴포스트(WP), 뉴욕타임즈(NYT) 등 전 세계 언론 및 방송사들은 취재 데이터 분석, 기사문장 생성, 뉴스제작에 이르기까지 다양한 방면에서 AI를 적극 활용 중이다. 이에 본 논문은 국내외 주요 미디어AI 서비스 유형과 특징을 속도와 규모, 다양성, 가치향상, 정확성 측면에서 종합적으로 분석하여 AI 저널리즘의 영향력과 발전 가능성을 평가한다. 나아가 균형 잡힌 시각을 유지하며 AI 도입의 기술적, 경영적, 법적 주요 쟁점들을 파악하고, 알고리즘 편향과 필터버블 등 첨단기술의 발전이 저널리즘 영역에 가져오는 도전을 체계적으로 준비하고자 한다. 마지막으로 AI와 미디어 산업의 상호지향적인 발전 방향을 모색하기 위해 사회적 합의를 통한 전향적 AI리터러시 원칙과 윤리적 가이드라인 개선의 필요성을 제언하며, 저널리즘의 본질적 가치와 임무를 조망한다.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• 제11권6호
• /
• pp.3029-3045
• /
• 2017

The main processes of a cognitive radio system include spectrum sensing, spectrum decision, spectrum sharing, and spectrum conversion. Experimental results show that these stages introduce a time delay that affects the spectrum sensing accuracy, reducing its efficiency. To reduce the time delay, the frequency spectrum prediction was proposed to alleviate the burden on the spectrum sensing. In this paper, the deep recurrent neural network (DRNN) was proposed to predict the spectrum of multiple time slots, since the existing methods only predict the spectrum of one time slot. The continuous state of a channel is divided into a many time slots, forming a time series of the channel state. Since there are more hidden layers in the DRNN than in the RNN, the DRNN has fading memory in its bottom layer as well as in the past input. In addition, the extended Kalman filter was used to train the DRNN, which overcomes the problem of slow convergence and the vanishing gradient of the gradient descent method. The spectrum prediction based on the DRNN was verified with a WiFi signal, and the error of the prediction was analyzed. The simulation results proved that the multiple slot spectrum prediction improved the spectrum efficiency and reduced the energy consumption of spectrum sensing.

• 한국군사과학기술학회지
• /
• 제19권6호
• /
• pp.690-697
• /
• 2016

EOIS(electro-optical imaging system) is the main target of the laser weapon. Specially, the image sensor will be vulnerable because EOIS focuses the incident laser beam onto the image sensor. Accordingly, the laser-induced damage of the image sensor needs to be identified for the counter-measure against the laser attack. In this study, the laser-induced damage of the CCD image sensor irradiated by the CW(continuous wave) NIR(near infrared) laser was experimentally investigated and mechanisms of those damage occurrences were analyzed. In the experiment, the near infrared CW fiber laser was used as a laser source. As the fluence, which is the product of the irradiance and the irradiation time, increased, the permanent damages such as discoloration and breakdown appeared sequentially. The discoloration occurred when the color filter was damaged and then the breakdown occurred when the photodiode and substrate were damaged. From the experimental results, LIDTs(laser-induced damage thresholds) of damages were roughly determined.

• Journal of information and communication convergence engineering
• /
• 제14권1호
• /
• pp.51-56
• /
• 2016

In this paper, we present a three-dimensional (3-D) automatic target recognition system based on optical integral imaging reconstruction. In integral imaging, elemental images of the reference and target 3-D objects are obtained through a lenslet array or a camera array. Then, reconstructed 3-D images at various reconstruction depths can be optically generated on the output plane by back-projecting these elemental images onto a display panel. 3-D automatic target recognition can be implemented using computational integral imaging reconstruction and digital nonlinear correlation filters. However, these methods require non-trivial computation time for reconstruction and recognition. Instead, we implement 3-D automatic target recognition using optical cross-correlation between the reconstructed 3-D reference and target images at the same reconstruction depth. Our method depends on an all-optical structure to realize a real-time 3-D automatic target recognition system. In addition, we use a nonlinear correlation filter to improve recognition performance. To prove our proposed method, we carry out the optical experiments and report recognition results.

• Journal of information and communication convergence engineering
• /
• 제8권1호
• /
• pp.51-58
• /
• 2010

This paper proposes an edge adaptive color interpolation for an ultra-small HD-grade complementary metal-oxide semiconductor (CMOS) video sensor in camera phones that can process 720-p/30-fps videos. Recently, proposed methods with great image quality perceptually reconstruct the green component and then estimate the red/blue component using the reconstructed green and neighbor red and blue pixels. However, these methods require the bulky memory line buffers in order to temporally store the reconstructed green components. The edge adaptive color interpolation method uses seven or nine patterns to calculate the six edge directions. At the same time, the threshold values are adaptively adjusted by the sum of the color values of the selected pixels. This method selects the suitable one among the patterns using two flowcharts proposed in this paper, and then interpolates the missing color values. For verification, we calculated the peak-signal-to-noise-ratio (PSNR) in the test images, which were processed by the proposed algorithm, and compared the calculated PSNR of the existing methods. The proposed color interpolation is also fabricated with the 0.18-${\mu}m$ CMOS flash memory process.

• 조명전기설비학회논문지
• /
• 제27권2호
• /
• pp.69-76
• /
• 2013

In this paper, electronic ballast using resonant inverter for HID lamp is designed and implemented. The proposed electronic ballast is used the soft switching technology ZVS(Zero Voltage Switching) to reduce turn-on and turn-off loss. The ignition of proposed electronic ballast is achieved by controlling a full bridge inverter which is consisted of LC filter for resonance. After ignition the ballast operates as a low frequency square wave inverter by controlling a full bridge inverter as a buck converter. After ignition at resonant frequency of $f_o$=160kHz, the switching frequency of a buck converter is consisted of 50kHz of high frequency and 170Hz of low frequency. This is for attenuating high frequency harmonics and avoiding acoustic resonance. The experimental results show that electronic ballast using resonant inverter is operated stably.

• Journal of the Optical Society of Korea
• /
• 제19권1호
• /
• pp.102-111
• /
• 2015

This paper proposes an algorithm for recovering the colors of objects from multiple near-infrared (near-IR) images. The International Commission on Illumination (CIE) color coordinates of objects are recovered from a series of gray images captured under multiple spectral near-IR illuminations using polynomial regression. The feasibility of the proposed algorithm is tested experimentally by using 24 color patches of the Color Rendition Chart. The experimental apparatus is composed of a monochrome digital camera without an IR cut-off filter and a custom-designed LED illuminator emitting multiple spectral near-IR illuminations, with peak wavelengths near the red edge of the visible band, namely at 700, 740, 780, and 860 nm. The average color difference between the original and the recovered colors for all 24 patches was found to be 11.1. However, if some particular patches with high value are disregarded, the average color difference is reduced to 4.2, and this value is within the acceptability tolerance for complex image on the display.