• Journal of the Architectural Institute of Korea Structure & Construction
• /
• v.36 no.3
• /
• pp.113-120
• /
• 2020

The purpose of this study is to compare the models of Deep Learning-based Convolution Neural Network(CNN) for concrete crack detection. The comparison models are AlexNet, GoogLeNet, VGG16, VGG19, ResNet-18, ResNet-50, ResNet-101, and SqueezeNet which won ImageNet Large Scale Visual Recognition Challenge(ILSVRC). To train, validate and test these models, we constructed 3000 training data and 12000 validation data with 256×256 pixel resolution consisting of cracked and non-cracked images, and constructed 5 test data with 4160×3120 pixel resolution consisting of concrete images with crack. In order to increase the efficiency of the training, transfer learning was performed by taking the weight from the pre-trained network supported by MATLAB. From the trained network, the validation data is classified into crack image and non-crack image, yielding True Positive (TP), True Negative (TN), False Positive (FP), False Negative (FN), and 6 performance indicators, False Negative Rate (FNR), False Positive Rate (FPR), Error Rate, Recall, Precision, Accuracy were calculated. The test image was scanned twice with a sliding window of 256×256 pixel resolution to classify the cracks, resulting in a crack map. From the comparison of the performance indicators and the crack map, it was concluded that VGG16 and VGG19 were the most suitable for detecting concrete cracks.

• Journal of Satellite, Information and Communications
• /
• v.7 no.3
• /
• pp.30-34
• /
• 2012

National Meteorological Satellite Center(NMSC) of Korea Meteorological Administration(KMA) has launched to implement the application development to get prepared for the space weather operation since 2010. As a action of KMA's space weather work, NMSC constructed Global Navigation Satellite System(GNSS) application system for meteorology and space weather. We will introduce NMSC's space weather application system which derives regional TEC(Total Electron Content) in near real time using nation-wide GNSS network data. First, We constructed system for collecting GNSS data, which is currently collecting about 80 stations operated by agencies like NGII(National Geographic Information Institute), Central Office of DGPS(Differential GPS), and KASI(Korea Astronomy and Space Science) including KMA's own data of 2 stations. In order to retreive regional TEC over Korean peninsular, we build up the automatic processes running every 1-hour. In these processes, firstly, GNSS data of every stations with 24 hours time window are processed to derive DCBs(Differential Code Biases) of each GNSS station and TEC values on every ionosphere piercing point(IPP). Then we made gridded regional TEC map with resolution of 0.25 degree from 31N, 121E to 41N, 135E by combination of all station results within 30 minutes window with assumption that TEC of a given point during a given 30 minutes window would have a constant value. The grid points without TEC value are interpolated using Barnes objective analysis. We presentour regional TEC maps, which can describe better on the status of ionosphere over Korean peninsular compared to IGS TEC maps.

• Journal of The Korean Astronomical Society
• /
• v.37 no.5
• /
• pp.349-353
• /
• 2004

Magnetic fields are an important ingredient of galaxy clusters and are indirectly observed on cluster scales as radio haloes and radio relics. One promising method to shed light on the properties of cluster wide magnetic fields is the analysis of Faraday rotation maps of extended extragalactic radio sources. We developed a Fourier analysis for such Faraday rotation maps in order to determine the magnetic power spectra of cluster fields. In an advanced step, here we apply a Bayesian maximum likelihood method to the RM map of the north lobe of Hydra A on the basis of our Fourier analysis and derive the power spectrum of the cluster magnetic field. For Hydra A, we measure a spectral index of -5/3 over at least one order of magnitude implying Kolmogorov type turbulence. We find a dominant scale of about 3 kpc on which the magnetic power is concentrated, since the magnetic autocorrelation length is ${\lambda}_B = 3 {\pm} 0.5\;kpc$. Furthermore, we investigate the influences of the assumption about the sampling volume (described by a window function) on the magnetic power spectrum. The central magnetic field strength was determined to be ${\~}7{\pm}2{\mu}G$ for the most likely geometries.

• Proceedings of the KSRS Conference
• /
• 2005.10a
• /
• pp.698-701
• /
• 2005

This paper examines that is extracted certain information in forest areas within high resolution imagery based on wavelet transformation. First of all, study areas are selected one more species distributed spots refer to forest type map. Next, study area is cut 256 x 256 pixels size because of image processing problem in large volume data. Prior to wavelet transformation, five texture parameters (contrast, dissimilarity, entropy, homogeneity, Angular Second Moment (ASM≫ calculated by using Gray Level Co-occurrence Matrix (GLCM). Five texture images are set that shifting window size is 3x3, distance .is 1 pixel, and angle is 45 degrees used. Wavelet function is selected Daubechies 4 wavelet basis functions. Result is summarized 3 points; First, Wavelet transformation images derived from contrast, dissimilarity (texture parameters) have on effect on edge elements detection and will have probability used forest road detection. Second, Wavelet fusion images derived from texture parameters and original image can apply to forest area classification because of clustering in Homogeneous forest type structure. Third, for grading evaluation in forest fire damaged area, if data fusion of established classification method, GLCM texture extraction concept and wavelet transformation technique effectively applied forest areas (also other areas), will obtain high accuracy result.

• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.10 no.10
• /
• pp.1829-1834
• /
• 2006

In this raper in order to recover sharp object boundaries we propose a new efficient stereo matching algorithm in which window size is varied to the distance from the boundaries of object. To this end, the processing region is divided into small subregions with a same area and the disparities of the center pixels in the subregions are calculated using a area-based algorithm with multiple windows. From the this disparity map we can find the edges of the contracted objects. The disparities of original image are obtained using the gradient constraint that means the disparity of the center pixel is similar to the ones of the remaining pixels in the subregion. from the experimental results it is found that the proposed algorithm is very good for recovering sharp object boundaries compared to the similar different algorithm.

• Journal of The Korean Society of Agricultural Engineers
• /
• v.57 no.2
• /
• pp.15-26
• /
• 2015

The purpose of this paper is to develop a software tool, PGA-CC (Projection of hydrology via Grid-based Assessment for Climate Change) to evaluate the present hydrologic cycle and the future watershed hydrology by climate change. PGA-CC is composed of grid-based input data pre-processing module, hydrologic cycle calculation module, output analysis module, and output data post-processing module. The grid-based hydrological model was coded by Fortran and compiled using Compaq Fortran 6.6c, and the Graphic User Interface was developed by using Visual C#. Other most elements viz. Table and Graph, and GIS functions were implemented by MapWindow. The applicability of PGA-CC was tested by assessing the future hydrology of South Korea by HadCM3 SRES B1 and A2 climate change scenarios. For the whole country, the tool successfully assessed the future hydrological components including input data and evapotranspiration, soil moisture, surface runoff, lateral flow, base flow etc. From the spatial outputs, we could understand the hydrological changes both seasonally and regionally.

• Journal of the Korean earth science society
• /
• v.36 no.5
• /
• pp.437-446
• /
• 2015

Many edge detection methods, based on horizontal and vertical derivatives, have been introduced to provide us with intuitive information about the horizontal distribution of a subsurface anomalous body. Understanding the characteristics of each edge detection method is important for selecting an optimized method. In order to compare the characteristics of the individual methods, this study applied each method to synthetic magnetic data created using homogeneous prisms with different sizes, the numbers of bodies, and spacings between them. Seven edge detection methods were comprehensively and quantitatively analyzed: the total horizontal derivative (HD), the vertical derivative (VD), the 3D analytic signal (AS), the title derivative (TD), the theta map (TM), the horizontal derivative of tilt angle (HTD), and the normalized total horizontal derivative (NHD). HD and VD showed average good performance for a single-body model, but failed to detect multiple bodies. AS traced the edge for a single-body model comparatively well, but it was unable to detect an angulated corner and multiple bodies at the same time. TD and TM performed well in delineating the edges of shallower and larger bodies, but they showed relatively poor performance for deeper and smaller bodies. In contrast, they had a significant advantage in detecting the edges of multiple bodies. HTD showed poor performance in tracing close bodies since it was sensitive to an interference effect. NHD showed great performance under an appropriate window.

• Journal of Multimedia Information System
• /
• v.7 no.1
• /
• pp.11-16
• /
• 2020

Deep Learning has become the most important technology in the field of artificial intelligence machine learning, with its high performance overwhelming existing methods in various applications. In this paper, an interactive window service based on object recognition technology is proposed. The main goal is to implement an object recognition technology using this deep learning technology to remove the existing eye tracking technology, which requires users to wear eye tracking devices themselves, and to implement an eye tracking technology that uses only usual cameras to track users' eye. We design an interactive system based on efficient eye detection and pupil tracking method that can verify the user's eye movement. To estimate the view-direction of user's eye, we initialize to make the reference (origin) coordinate. Then the view direction is estimated from the extracted eye pupils from the origin coordinate. Also, we propose a blink detection technique based on the eye apply ratio (EAR). With the extracted view direction and eye action, we provide some augmented information of interest without the existing complex and expensive eye-tracking systems with various service topics and situations. For verification, the user guiding service is implemented as a proto-type model with the school map to inform the location information of the desired location or building.

• Proceedings of the Korean Institute of Navigation and Port Research Conference
• /
• 2002.11a
• /
• pp.1-6
• /
• 2002

To provide an easy operation of drift prediction model in SAR(search and rescue) mission a GUI program running on Window environment has developed. Users can make choice of input data on the screen by just clicking the mouse and the prediction results of datum points and trajectories of vessels are drawn on the map. The program contains both Leeway Equation model and mathematical model. The FORTRAN language was used in programming and Lehay Winteracter 4.0 software was utilized for graphic presentation. The result of May, 2001 Busan field experiment was plotted with that of model prediction for demonstration purpose.

• Proceedings of the Korean Society of Broadcast Engineers Conference
• /
• 2012.07a
• /
• pp.83-86
• /
• 2012

본 논문에서는 다양한 변이 추정 방식 중 영역기반(Area-based) 알고리듬과 특정기반(Feature-based) 알고리듬을 결합한 하이브리드(Hybrid) 변이추정 알고리듬을 제안한다. 제안하는 알고리듬은 Features from Accelerated Segment Test(FAST) 코너 점 추출기[2]를 이용하여 좌, 우 영상 각각의 특징 점을 추출한 후, 특징 점들의 정보를 이용한 스테레오 정함을 통해 신뢰도 높은 초기 변이지도(Disparity map)를 생생하게 된다. 그러나 생성된 초기 변이지도는 조밀하지 못하므로, 조밀한 변이 지도를 획득하기 위해 특징점이 추출된 영역에 대해서는 추정된 초기 변이 값을 이웃 픽셀과의 색 유사도를 고려하여 전파시키고 특징 점이 추출되지 않은 영역에 대해서는 이진 윈도우(Binary window)를 활용한 영역기반 변이추정 알고리듬[1]을 이용하여 변이 값을 추정한다. 이를 통해, 제안 알고리듬은 특징 기반 알고리듬에서 발생할 수 있는 보간법 문제를 해결함과 동시에 신뢰도가 높은 초기 변이지도를 사용함으로써, 영역 기반 알고리듬의 정합 오차를 줄여 신뢰도 높은 변이지도를 생생할 수 있다. 실험 결과 추정된 초기 변이지도는 ground truth와 비교 시 약 99%이상의 정확도를 보이며, 특징 점이 추출된 영역에서 기존의 영역기반 알고리듬보다 더 정확한 변이 값이 추정되었음을 확인하였다.