• Journal of the Korean Society of Surveying, Geodesy, Photogrammetry and Cartography
• /
• v.36 no.5
• /
• pp.423-432
• /
• 2018

Surface water mapping has been widely used in various remote sensing applications. Water indices have been commonly used to distinguish water bodies from land; however, determining the optimal threshold and discriminating water bodies from similar objects such as shadows and snow is difficult. Deep learning algorithms have greatly advanced image segmentation and classification. In particular, FCN (Fully Convolutional Network) is state-of-the-art in per-pixel image segmentation and are used in most benchmarks such as PASCAL VOC2012 and Microsoft COCO (Common Objects in Context). However, these data sets are designed for daily scenarios and a few studies have conducted on applications of FCN using large scale remotely sensed data set. This paper aims to fine-tune the pre-trained FCN network using the CRMS (Coastwide Reference Monitoring System) data set for surface water mapping. The CRMS provides color infrared aerial photos and ground truth maps for the monitoring and restoration of wetlands in Louisiana, USA. To effectively learn the characteristics of surface water, we used pre-trained the DeepWaterMap network, which classifies water, land, snow, ice, clouds, and shadows using Landsat satellite images. Furthermore, the DeepWaterMap network was fine-tuned for the CRMS data set using two classes: water and land. The fine-tuned network finally classifies surface water without any additional learning process. The experimental results show that the proposed method enables high-quality surface mapping from CRMS data set and show the suitability of pre-trained FCN networks using remote sensing data for surface water mapping.

• Journal of the Institute of Convergence Signal Processing
• /
• v.10 no.3
• /
• pp.164-169
• /
• 2009

This paper proposes a road detection method using BP(Back-Propagation) neural network based on texture information of the each candidate road region segmented for satellite images. To segment the candidate road regions, the histogram-based binarization method proposed by N.Otsu is firstly performed and the neighboring regions surrounding road regions are then removed. And after extracting the principal color using the histogram of the segmented foreground, the candidate road regions are classified into the regions within ${\pm}25$ of the principal color. Finally, the road regions are segmented using BP neural network based on texture information of the candidate regions. The texture information in this paper is calculated using co-occurrence matrix and is used as an input data of the BP neural network. The proposed method is based on the fact that the road has the constant intensity and shape. The experiment demonstrated the validity of the proposed method and showed 90% detection accuracy for the various images.

• Journal of information and communication convergence engineering
• /
• v.4 no.1
• /
• pp.18-22
• /
• 2006

A small-sized gas identification system has been fabricated and characterized using an integrated gas sensor array and artificial neural-network. The sensor array consists of four thick-film oxide semiconductor gas sensors whose sensing layers are $In_{2}O_{3}-Sb_{2}O_{5}-Pd-doped\;SnO_2$ + Pd-coated layer, $La_{2}O_{5}-PdCl_{2}-doped\;SnO_2,\;WO_{3}-doped\;SnO_{2}$ + Pt-coated layer and $ThO_{2}-V_{2}O_{5}-PdCl_{2}\;doped\;SnO_{2}$. The small-sized gas identification instrument is composed of a GMS 81504 containing an internal ROM (4k bytes), a RAM (128 bytes) and four-channel AD converter as MPU, LEDs for displaying alarm conditions for three gases (liquefied petroleum gas: LPG, liquefied natural gas: LNG and carbon monoxide: CO) and interface circuits for them. The instrument has been used to identify alarm conditions for three gases among the real circumstances and the identification has been successfully demonstrated.

• The Journal of Korean Institute of Communications and Information Sciences
• /
• v.30 no.1A
• /
• pp.70-75
• /
• 2005

The efficient hardware design of the the algorithm is important in wide variety of DSP. One example is OFDM(Orthogonal Frequency Division Multiplexing) based WLAN(Wireless Local Area Network) systems which place high requirements on throughput and power consumption on FFT. The output RAM is composed of two banks of $64{\times}W.$ The banks are swapped immediately following the falling edge or the start signal strobe. This bank swapping allows 64-Point FFT to continue Processing samples and to continue filling the alternative bank, without affecting the data flow outputs.

• Communications for Statistical Applications and Methods
• /
• v.31 no.2
• /
• pp.213-234
• /
• 2024

In recent decades, increasing research attention has been directed toward predicting the price of stocks in financial markets using deep learning methods. For instance, recurrent neural network (RNN) is known to be competitive for datasets with time-series data. Long short term memory (LSTM) further improves RNN by providing an alternative approach to the gradient loss problem. LSTM has its own advantage in predictive accuracy by retaining memory for a longer time. In this paper, we combine both supervised and unsupervised dimension reduction methods with LSTM to enhance the forecasting performance and refer to this as a dimension reduction based LSTM (DR-LSTM) approach. For a supervised dimension reduction method, we use methods such as sliced inverse regression (SIR), sparse SIR, and kernel SIR. Furthermore, principal component analysis (PCA), sparse PCA, and kernel PCA are used as unsupervised dimension reduction methods. Using datasets of real stock market index (S&P 500, STOXX Europe 600, and KOSPI), we present a comparative study on predictive accuracy between six DR-LSTM methods and time series modeling.

• KIPS Transactions on Computer and Communication Systems
• /
• v.4 no.1
• /
• pp.11-14
• /
• 2015

WSN (Wireless Sensor Network) based on low-power is one of the core technologies in the ubiquitous society. In this paper, we present a border surveillance and control system in WSN environment. The system consists of static sensor node, mobile sensor node, static gateway, mobile gateway, server and mobile application. Mobile applications are divided into user mode and manager mode. So users monitor border surveillance through mobile phone and get information of border network environment without time and space constraints. In manager mode, for the flexible operation of nodes, manager can update to the software remotely and adjust the position of the mobile node. And also we implement a suitable multi-hop routing protocol for scalable low-power sensor nodes and confirm that the system operates well in WSN environment.

• The Journal of the Korea institute of electronic communication sciences
• /
• v.15 no.6
• /
• pp.1081-1088
• /
• 2020

In this paper, the method of detecting sea fog through CCTV image is proposed based on convolutional neural networks. The study data randomly extracted 1,0004 images, sea-fog and not sea-fog, from a total of 11 ports or beaches (Busan Port, Busan New Port, Pyeongtaek Port, Incheon Port, Gunsan Port, Daesan Port, Mokpo Port, Yeosu Gwangyang Port, Ulsan Port, Pohang Port, and Haeundae Beach) based on 1km of visibility. 80% of the total 1,0004 datasets were extracted and used for learning the convolutional neural network model. The model has 16 convolutional layers and 3 fully connected layers, and a convolutional neural network that performs Softmax classification in the last fully connected layer is used. Model accuracy evaluation was performed using the remaining 20%, and the accuracy evaluation result showed a classification accuracy of about 96%.

• Proceedings of the Korean Operations and Management Science Society Conference
• /
• 2007.11a
• /
• pp.16-19
• /
• 2007

At an integrated steel mill, raw materials such as coal and iron ore are discharged by ships through multiple unloaders. The discharged raw material is then transported to storage yards through multiple routes established simultaneously on a fairly complicated belt conveyer network. Formulating an efficient unloading schedule is a quite cumbersome task due to the insufficient number of berths and unloaders as well as the potential conflict among routes being used simultaneously. In this paper, we propose a solution approach to the scheduling problem and describe the prototype system that we built as an implementation of our approach.

• Journal of Electrical Engineering and Technology
• /
• v.12 no.4
• /
• pp.1386-1396
• /
• 2017

In this paper, a new approach is proposed to select the optimal sitting and sizing of distributed solar photovoltaic generation (SPVG) in a radial electrical distribution systems (EDS) considering load/generation uncertainties. Here, distributed generations (DGs) allocation problem is modeled as optimization problem with network loss based objective function under various equality and inequality constrains in an uncertain environment. A boundary power flow is utilized to address the uncertainties in load/generation forecasts. This approach facilitates the consideration of random uncertainties in forecast having no statistical history. Uncertain solar irradiance is modeled by beta distribution function (BDF). The resulted optimization problem is solved by a new Dynamic Harmony Search Algorithm (DHSA). Dynamic band width (DBW) based DHSA is proposed to enhance the search space and dynamically adjust the exploitation near the optimal solution. Proposed approach is demonstrated for two standard IEEE radial distribution systems under different scenarios.

• Proceedings of the Korean Information Science Society Conference
• /
• 2007.10c
• /
• pp.394-397
• /
• 2007

본 논문에서는 문자 인식의 특징 선별 방법으로 중간점 알고리즘을 이용하는 방법을 제안한다. 영상자료의 특징들로부터 중간점을 선별하고 심볼패턴을 이용하여 필기체 문자를 인식한다. 이 방법은 사전에 많은 심볼 패턴을 학습해야 하지만 한글과 영어의 높은 인식률을 보이고 있으며, 특히 복잡한 문자들의 경우 좋은 결과를 낸다. 여기서는 중간점 알고리즘으로 입력된 데이터를 심볼 패턴과 비교하고, 심볼 영역에 의해 최적 판별 기저를 탐색한 후, 그것을 특징으로 선택한다. 또한 사전 기능과 투명도 기능을 구현하여 필기체 인식을 이용한 여러 활용 방안을 제시한다.