• The Korean Journal of Applied Statistics
• /
• v.36 no.2
• /
• pp.101-114
• /
• 2023

The data augmentation technique is effectively used to solve the problem of overfitting the model by allowing the training dataset to be viewed from various perspectives. In addition to image augmentation techniques such as rotation, cropping, horizontal flip, and vertical flip, occlusion-based data augmentation methods such as Cutmix and Cutout have been proposed. For models based on speech data, it is possible to use an occlusion-based data-based augmentation technique after converting a 1D speech signal into a 2D spectrogram. In particular, SpecAugment is an occlusion-based augmentation technique for speech spectrograms. In this study, we intend to compare and study data augmentation techniques that can be used in the problem of false-voice detection. Using data from the ASVspoof2017 and ASVspoof2019 competitions held to detect fake audio, a dataset applied with Cutout, Cutmix, and SpecAugment, an occlusion-based data augmentation method, was trained through an LCNN model. All three augmentation techniques, Cutout, Cutmix, and SpecAugment, generally improved the performance of the model. In ASVspoof2017, Cutmix, in ASVspoof2019 LA, Mixup, and in ASVspoof2019 PA, SpecAugment showed the best performance. In addition, increasing the number of masks for SpecAugment helps to improve performance. In conclusion, it is understood that the appropriate augmentation technique differs depending on the situation and data.

• Journal of Acupuncture Research
• /
• v.17 no.4
• /
• pp.79-87
• /
• 2000

Objectives : To evaluate the findings useful for differential diagnosis and associated abnormaiities of isthmic spondylolisthesis and degenerative spondylolisthesis on CT. Materials and methods : We reviewed retrospectively the CT images of 65 patients who were diagnosed spondylolisthesis during 3 years period. Our technique was 5mm slices at 5mm intervals with gantry angle to parallel the interspaces. Also reformatted sagittal views were taken. 41 patients were isthmic spondylolisthesis and 24 patients were degenerative spondylolisthesis. Resuits : Isthmic spondylolisthesis. 1. Isthmic type was more common at L5-S1. 2. The degree of anterior displacement was grade I and II. 3. The plane of defect was more horizontal than the usual facet joint. 4. The defect had an irregular shape. 5. Medial aspect of bone just anterior to defect had a small round prominence. 6. Anteroposterior elongation of the spinal canal was common. 7. Pseudobulging disk was common. 8. The most common associated abnormality was a HNP at the upper level of the defect. Degenerative spondylolisthesis. 1. Degenerative type was more common at L4-5. 2. The degree of anterior disptacement was grade I and II. 3. The Plane of facet joint was oriented obliquely instead of horizontally. 4. The posterior facet(inferior facet of superior vertebra) was anteriorly displaced. 5. Bony spur of the posterior portion of anterior facet was seen. 6. The facet joints often contain gas(vaccum phenomenum). 7. The most common associated abnormality was a HNP at the level of the displacement. Conclusions : CT is a highly accurate and most sensitive technique for recognition, differential diagnosis of isthmic and degenerative types and the detection of associated abnormalities.

• Geophysics and Geophysical Exploration
• /
• v.13 no.1
• /
• pp.96-101
• /
• 2010

A pulse radar system has been developed recently to detect dormant underground tunnels that are deeply located at depths of hundreds of metres. To check the ability of the radar system to detect an obliquely oriented tunnel, five different borehole pairs in the tunnel test site were chosen so that the horizontal lines-of-sight cut the tunnel axis obliquely, in $15^{\circ}$ steps. The pulse radar signatures were measured over a depth range of 20 m around the centre of the air-filled tunnel. Three canonical parameters, consisting of the arrival time, attenuation, and dispersion time were extracted from the first and second peaks of the measured radar signatures. Using those parameters, the radar system can detect obliquely oriented tunnels at various angles up to 45 from the transmitter-receiver line of sight.

• Journal of Digital Contents Society
• /
• v.15 no.2
• /
• pp.209-217
• /
• 2014

It is very important to classify a lot of table-form documents into the same type of classes or to extract information filled in the template automatically. For these, it is necessary to accurately analyze table-form structure. This paper proposes an algorithm to extract corner points based on line edge segments and to classify the type of junction from table-form images. The algorithm preprocesses image through binarization, skew correction, deletion of isolated small area of black color because that they are probably generated by noises.. And then, it processes detections of edge block, line edges from a edge block, corner points. The extracted corner points are classified as 9 types of junction based on the combination of horizontal/vertical line edge segments in a block. The proposed method is applied to the several unconstraint document images such as tax form, transaction receipt, ordinary document containing tables, etc. The experimental results show that the performance of point detection is over 99%. Considering that almost corner points make a correspondence pair in the table, the information of type of corner and width of line may be useful to analyse the structure of table-form document.

• Journal of Cadastre & Land InformatiX
• /
• v.45 no.2
• /
• pp.117-130
• /
• 2015

A least squares method for adjusting the horizontal network satisfies the conditions which is minimizing the sum of the squares of errors based on probability theory. This research compared accuracy of 3rd cadastral control points adjusted by traditional and least square method with respect to the result of Network-RTK. Test results showed the least square method more evenly distribute closure error than traditional method. Mean errors of least square and traditional adjusting method are 2.7cm, 2.2cm respectively. In addition, blunder in angle observations can be detected by comparing position errors which calculated by forward and backward initial coordinates. However, distance blunder cannot offer specific observation line occurred mistake because distance error propagates several observation lines which have similar directions.

• Journal of the Korean Society of Surveying, Geodesy, Photogrammetry and Cartography
• /
• v.33 no.1
• /
• pp.23-30
• /
• 2015

A shoreline mapping is essential for describing coastal areas, estimating coastal erosions and managing coastal properties. This study has planned to map the 3D shorelines with the airborne LiDAR(Light Detection and Ranging) data and the KOMPSAT-2 imagery, acquired in Uljin, Korea. Following to the study, the DSM(Digital Surface Model) is generated firstly with the given LiDAR data, while the NDWI(Normalized Difference Water Index) imagery is generated by the given KOMPSAT-2 imagery. The classification method is employed to generate water and land clusters from the NDWI imagery, as the 2D shorelines are selected from the boundaries between the two clusters. Lastly, the 3D shorelines are constructed by adding the elevation information obtained from the DSM into the generated 2D shorelines. As a result, the constructed 3D shorelines have had 0.90m horizontal accuracy and 0.10m vertical accuracy. This statistical results could be concluded in that the generated 3D shorelines shows the relatively high accuracy on classified water and land surfaces, but relatively low accuracies on unclassified water and land surfaces.

• Transactions of the Korean Society of Mechanical Engineers
• /
• v.11 no.1
• /
• pp.118-123
• /
• 1987

In relation to the dust detonatians which have imposed severe damages on the industry, the ignitability of various dusts has been investigated on a horizontal shock tube in this study. By using a newly designed air injector, very well-distributed clouds could be obtained. The proper reflected shock conditions have been generated by placing a reflector 1.5cm behind the air injector, which reflected the incident shock wave. The incident shock waves in the range of Mach number 2.8-3.3 created the postreflected shock temperature of 1200-1600K. Experimentally the ignition delay was defined as the time interval between the arrival of a reflected shock wave at dusts and the detection of visible light. Measured ignition delays of dusts investigated were located lower than 1msec under the above conditions. These values are one-order higher than those in the incident shock wave condition. In this type of ignitiion process the following three processes are considered to play important roles; heating of a particle, generation of volatile gas by endothermic devolatilization process, and its diffusion from the particle surface and the formation of stoichiometric mixture with oxidizer.

• Atmosphere
• /
• v.19 no.3
• /
• pp.269-287
• /
• 2009

CAT (clear-air turbulence) forecasting algorithm, the Graphical Turbulence Guidance (GTG) system developed at NCAR (national center for atmospheric research), is evaluated with available observations (e.g., pilot reports; PIREPs) reported in South Korea during the recent 5 years (2003-2008, excluding 2005). The GTG system includes several steps. First, 44 CAT indices are calculated in the domain of the Regional Data Assimilation and Prediction System (RDAPS) analysis data with 30 km horizontal grid spacing provided by KMA (Korean Meteorological Administration). Second, 10 indices that performed ten best forecasting scores are selected. Finally, 10 indices are combined by measuring the score based on the probability of detection, which is calculated using PIREPs exclusively of moderate-or-greater intensity. In order to investigate the best performance of the GTG system in Korea, various statistical examinations and sensitivity tests of the GTG system are performed by yearly and seasonally classified PIREPs. Performances of the GTG system based on yearly distributed PIREPs have annual variations because the compositions of indices are different from each year. Seasonal forecasting is generally better than yearly forecasting, because selected CAT indices in each season represent meteorological condition much more properly than applying the selected CAT indices to all seasons. Wintertime forecasting is the best among the four seasonal forecastings. This is likely due to that the GTG system consists of many CAT indices related to the jet stream, and turbulence associated with the jet stream can be activated mostly in wintertime under strong jet magnitude. On the other hand, summertime forecasting skill is much less than other seasons. Compared with current operational CAT prediction system (KITFA; Korean Integrated Turbulence Forecasting System), overall performance of the GTG system is better when CAT indices are selected seasonally.

• The Journal of the Korea Contents Association
• /
• v.2 no.2
• /
• pp.17-26
• /
• 2002

In this paper, we suggest a method to acquire the moving object centered video by panning and tilting a camera automatically according to motion vectors calculated by detecting the motion of a moving object on video steam. We create a difference image by estimating the intensity difference at the grid points of neighboring frames. And we detect the motion using both horizontal projection histogram and vertical projection histogram and decide the center of motion part. Then we calculate a new direction and degree of the motion by comparing coordinates at the center of current motion and the center of previous motion. By controling the RCM using these Motion vectors, we can get video stream positioned unwire object on the center of video frame. Through the experiments, we could get a moving object centered video stream continuously arid monitor remotely by implementing sever/client architecture based on the web.

• Proceedings of the KSRS Conference
• /
• v.2
• /
• pp.590-593
• /
• 2006

Timber inventory management includes to measure and update forest attributes, which is crucial information for private companies and public organizations in property assessment and environment monitoring. Field measurement would be accurate, but time-consuming and inefficient. For the reason, remote sensing technology has been an alternative to field measurement from an economic perspective. Among several sensors, LiDAR and Radar interferometry are known for their efficiency for forest monitoring because they are less influenced by weather and light conditions, and provide reasonably accurate vertical/horizontal measurement for a large area in a short period. For example, Shuttle Radar Topography Mission (SRTM) and National Elevation Dataset (NED) in the U.S. can provide tree height information and DSM. On the other hand, LiDAR DSM (the first return) and DEM (the last return) can also present tree height estimation. With respect to project site of loblolly pine plantation in Louisiana in the U.S., the accuracy of SRTM C-Band approach estimating tree height was assessed by the LiDAR approaches. In addition, SRTM X-Band and NED were also compared with the results. Plantation year in inventory GIS, which is directly related to forest age, is high correlated with the difference between SRTM C-Band and NED. As a byproduct, several stands of age mismatch could be recognized using an outlier detection algorithm, and optical satellite image (ETM+) were used to verify the mismatch. The findings of this study were (1) the confirmation of usefulness of the SRTM DSM for forest monitoring and (2) Multi-sensors- Radar, LiDAR, ETM+, MODIS can be used for accuracy improvement of forest inventory GIS altogether.