• Journal of Intelligence and Information Systems
• /
• v.26 no.2
• /
• pp.1-25
• /
• 2020

In this paper, we suggest an application system architecture which provides accurate, fast and efficient automatic gasometer reading function. The system captures gasometer image using mobile device camera, transmits the image to a cloud server on top of private LTE network, and analyzes the image to extract character information of device ID and gas usage amount by selective optical character recognition based on deep learning technology. In general, there are many types of character in an image and optical character recognition technology extracts all character information in an image. But some applications need to ignore non-of-interest types of character and only have to focus on some specific types of characters. For an example of the application, automatic gasometer reading system only need to extract device ID and gas usage amount character information from gasometer images to send bill to users. Non-of-interest character strings, such as device type, manufacturer, manufacturing date, specification and etc., are not valuable information to the application. Thus, the application have to analyze point of interest region and specific types of characters to extract valuable information only. We adopted CNN (Convolutional Neural Network) based object detection and CRNN (Convolutional Recurrent Neural Network) technology for selective optical character recognition which only analyze point of interest region for selective character information extraction. We build up 3 neural networks for the application system. The first is a convolutional neural network which detects point of interest region of gas usage amount and device ID information character strings, the second is another convolutional neural network which transforms spatial information of point of interest region to spatial sequential feature vectors, and the third is bi-directional long short term memory network which converts spatial sequential information to character strings using time-series analysis mapping from feature vectors to character strings. In this research, point of interest character strings are device ID and gas usage amount. Device ID consists of 12 arabic character strings and gas usage amount consists of 4 ~ 5 arabic character strings. All system components are implemented in Amazon Web Service Cloud with Intel Zeon E5-2686 v4 CPU and NVidia TESLA V100 GPU. The system architecture adopts master-lave processing structure for efficient and fast parallel processing coping with about 700,000 requests per day. Mobile device captures gasometer image and transmits to master process in AWS cloud. Master process runs on Intel Zeon CPU and pushes reading request from mobile device to an input queue with FIFO (First In First Out) structure. Slave process consists of 3 types of deep neural networks which conduct character recognition process and runs on NVidia GPU module. Slave process is always polling the input queue to get recognition request. If there are some requests from master process in the input queue, slave process converts the image in the input queue to device ID character string, gas usage amount character string and position information of the strings, returns the information to output queue, and switch to idle mode to poll the input queue. Master process gets final information form the output queue and delivers the information to the mobile device. We used total 27,120 gasometer images for training, validation and testing of 3 types of deep neural network. 22,985 images were used for training and validation, 4,135 images were used for testing. We randomly splitted 22,985 images with 8:2 ratio for training and validation respectively for each training epoch. 4,135 test image were categorized into 5 types (Normal, noise, reflex, scale and slant). Normal data is clean image data, noise means image with noise signal, relfex means image with light reflection in gasometer region, scale means images with small object size due to long-distance capturing and slant means images which is not horizontally flat. Final character string recognition accuracies for device ID and gas usage amount of normal data are 0.960 and 0.864 respectively.

• Proceedings of the Korea Water Resources Association Conference
• /
• 2020.06a
• /
• pp.167-167
• /
• 2020

Annual sediment deposition in reservoirs behind weirs poses flood risk, while its accurate prediction remains a challenge. Sediment management by hydraulic flushing is an effective method to maintain reservoir storage. In this study, an integrated approach to predict sediment inflow and sediment flushing simulation in reservoirs is presented. The annual sediment inflow prediction was carried out with Artificial Neural Networks (ANN) modelling. RESCON model was applied for quantification of sediment flushing feasibility criteria. The integrated approach was applied on Sangju Weir and also on estuary of Nakdong River (NREB). The mean annual sediment inflow predicted at Sangju Weir and NREB was 400,000 ㎥ and 170,000 ㎥, respectively. The sediment characteristics gathered were used to setup RESCON model and sediment balance ratio (SBR) and long term capacity ratio (LTCR) were used as flushing efficiency indicators. For Sangju Weir, the flushing discharge, Q_f = 140 ㎥/s with a drawdown of 5 m, and flushing duration, T_f = 10 days was necessary for efficient flushing. At NREB site, the parameters for efficient flushing were Q_f = 80 ㎥/s, T_f = 5 days, N = 1, El_f = 2.24 m. The hydraulic flushing was concluded feasible for sediment management at both Sangju Weir and NREB.

• Annual Conference on Human and Language Technology
• /
• 2018.10a
• /
• pp.527-529
• /
• 2018

초기 자연언어처리에 FNN(Feedforward Neural Network)을 적용한 연구들에 비해 LSTM(Long Short-Term Memory)은 현재 시점의 정보뿐만 아니라 이전 시점의 정보를 담고 있어 문장을 이루는 어절들, 어절을 이루는 형태소 등 순차적인(sequential) 데이터를 처리하는데 좋은 성능을 보인다. 본 논문에서는 스택과 버퍼에 있는 어절을 양방향 LSTM encoding을 이용한 representation으로 표현하여 전이기반 의존구문분석에 적용하여 현재 UAS 89.4%의 정확도를 보였고, 자질 추가 및 정제작업을 통해 성능이 개선될 것으로 보인다.

• Journal of the Korean Institute of Telematics and Electronics B
• /
• v.32B no.12
• /
• pp.1635-1643
• /
• 1995

This paper proposes an algorithm for machine recognition of phonemes in continuously spoken Korean. The proposed algorithm is a static strategy neural network. The algorithm uses, at the stage of training neurons, features such as the rate of zero crossing, short-term energy, and either PARCOR or auditory-like perceptual linear prediction(PLP) but not both, covering a time of 171ms long. Numerical results show that the algorithm with PLP achieves approximately the frame-based phoneme recognition rate of 99% for small vocabulary recognition experiments. Based on this it is concluded that the proposed algorithm with PLP analysis is effective in phoneme recognition.

• Proceedings of the Korean Society of Broadcast Engineers Conference
• /
• 2016.11a
• /
• pp.13-16
• /
• 2016

본 논문에서는 순환 신경망을 이용하여 동영상에서의 배경과 전경을 구분하는 알고리즘을 제안한다. 순환 신경망은 일련의 순차적인 입력에 대해서 내부의 루프(loop)를 통해 이전 입력에 의한 정보를 지속할 수 있도록 구성되는 신경망을 말한다. 순환 신경망의 여러 구조들 가운데, 우리는 장기적인 관계에도 반응할 수 있도록 장단기 기억 신경망(Long short-term memory networks, LSTM)을 사용했다. 그리고 동영상에서의 시간적인 연결 뿐 아니라 공간적인 연관성도 배경과 전경을 판단하는 것에 영향을 미치기 때문에, 공간적 순환 신경망을 적용하여 내부 신경망(hidden layer)들의 정보가 공간적으로 전달될 수 있도록 신경망을 구성하였다. 제안하는 알고리즘은 기본적인 배경차분 동영상에 대해 기존 알고리즘들과 비교할만한 결과를 보인다.

• Annual Conference on Human and Language Technology
• /
• 2015.10a
• /
• pp.85-89
• /
• 2015

본 논문에서는 시퀀스 레이블링 문제에 적합하다고 알려진 Long Short Term Memory Recurrent Neural Network에 아웃풋간의 의존관계를 추가한 LSTM-CRF(Conditional Random Field)를 이용하여 생명과학분야 개체명 인식 시스템을 구축하였다. 학습 및 평가를 위해 BioNLP 2011-st REL data를 개체명 인식 실험에 사용하였으며, 실험결과 LSTM-CRF를 사용한 시스템은 81.83의 F1-score를 기록해, 기존의 시스템인 "BANNER"의 F1-score 81.96과 비슷한 성능을 보였다.

• Annual Conference on Human and Language Technology
• /
• 2015.10a
• /
• pp.224-227
• /
• 2015

상호참조해결은 특정 개체에 대해 다르게 표현한 단어들을 서로 연관지어 주며, 이러한 개체에 대해 표현한 단어들을 멘션(mention)이라 하며, 이런 멘션을 찾아내는 것을 멘션탐지(mention detection)라 한다. 멘션은 명사나 명사구를 기반으로 정의되며, 명사구의 경우에는 수식어를 포함하기 때문에 멘션탐지를 순차 데이터 문제(sequence labeling problem)로 정의할 수 있다. 순차 데이터 문제에는 Recurrent Neural Network(RNN) 종류의 모델을 적용할 수 있으며, 모델들은 Long Short-Term Memory(LSTM) RNN, LSTM Recurrent CRF(LSTM-CRF), Bidirectional LSTM-CRF(Bi-LSTM-CRF) 등이 있다. LSTM-RNN은 기존 RNN의 그레디언트 소멸 문제(vanishing gradient problem)를 해결하였으며, LSTM-CRF는 출력 결과에 의존성을 부여하여 순차 데이터 문제에 더욱 최적화 하였다. Bi-LSTM-CRF는 과거입력자질과 미래입력자질을 함께 학습하는 방법으로 최근에 가장 좋은 성능을 보이고 있다. 이에 따라, 본 논문에서는 멘션탐지에 Bi-LSTM-CRF를 적용할 것을 제안하며, 각 딥 러닝 모델들에 대한 비교실험을 보인다.

• Journal of information and communication convergence engineering
• /
• v.9 no.4
• /
• pp.369-374
• /
• 2011

Generally the neural network and the Fuzzy compensative algorithms are applied to forecast the time series for power demand with the characteristics of a nonlinear dynamic system, but, relatively, they have a few prediction errors. They also make long term forecasts difficult because of sensitivity to the initial conditions. In this paper, we evaluate the chaotic characteristic of electrical power demand with qualitative and quantitative analysis methods and perform a forecast simulation of electrical power demand in regular sequence, attractor reconstruction and a time series forecast for multi dimension using Lyapunov Exponent (L.E.) quantitatively. We compare simulated results with previous methods and verify that the present method is more practical and effective than the previous methods. We also obtain the hourly predictability of time series for power demand using the L.E. and evaluate its accuracy.

• Proceedings of the Korea Water Resources Association Conference
• /
• 2019.05a
• /
• pp.188-188
• /
• 2019

본 연구에서는 텐서플로우를 이용한 관측자료 기반의 수위예측 연구를 수행하였다. 대상유역은 도림천 유역으로 선정하였으며 관측강우와 상류하천의 수위자료를 이용하여 하류인 도림교지점의 수위를 예측하였으며 다른 변수는 배제하였다. 사용된 모형은 시계열 데이터예측에 우수한 성능을 보이는 RNN(Recurrent Neural Network)과 LSTM(Long Short Term Memory networks)을 이용하였으며 수위자료는 2005년부터 2016년도 10분단위 관측강우와 수위 데이터를 학습하여 2017년도 수위데이터를 예측하도록 하였다. 본 연구를 통하여 홍수기 실시간 수위예측이 가능할것으로 판단되며 도시지역 골든타임 확보에 활용될 것으로 판단된다.

• New & Renewable Energy
• /
• v.18 no.2
• /
• pp.18-25
• /
• 2022

In this study, we propose a wind power generation prediction system that applies machine learning and data mining to predict wind power generation. This system increases the utilization rate of new and renewable energy sources. For time-series data, the data set was established by measuring wind speed, wind generation, and environmental factors influencing the wind speed. The data set was pre-processed so that it could be applied appropriately to the model. The prediction system applied the CNN (Convolutional Neural Network) to the data mining process and then used the LSTM (Long Short-Term Memory) to learn and make predictions. The preciseness of the proposed system is verified by comparing the prediction data with the actual data, according to the presence or absence of data mining in the model of the prediction system.