• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.26 no.9
• /
• pp.1330-1339
• /
• 2022

3D object detection generally aims to detect relatively large data such as automobiles, buses, persons, furniture, etc, so it is vulnerable to small object detection. In addition, in an environment with limited resources such as embedded devices, it is difficult to apply the model because of the huge amount of computation. In this paper, the accuracy of small object detection was improved by focusing on local features using only one layer, and the inference speed was improved through the proposed knowledge distillation method from large pre-trained network to small network and adaptive quantization method according to the parameter size. The proposed model was evaluated using SUN RGB-D Val and self-made apple tree data set. Finally, it achieved the accuracy performance of 62.04% at mAP@0.25 and 47.1% at mAP@0.5, and the inference speed was 120.5 scenes per sec, showing a fast real-time processing speed.

• Journal of Broadcast Engineering
• /
• v.2 no.2
• /
• pp.136-145
• /
• 1997

As a 3D mesh model consists of a lot of vertices and polygons and each vertex position is represented by three 32 bit floating-point numbers in a 3D coordinate, the amount of data needed for representing the model is very excessive. Thus, in order to store and/or transmit the 3D model efficiently, a 3D model compression is necessarily required. In this paper, a 3D model compression method using PRVQ (predictive residual vector quantization) is proposed. Its underlying idea is based on the characteristics such as high correlation between the neighboring vertex positions and the vectorial property inherent to a vertex position. Experimental results show that the proposed method obtains higher compression ratio than that of the existing methods and has the advantage of being capable of transmitting the vertex position data progressively.

• The Korean Journal of Applied Statistics
• /
• v.29 no.2
• /
• pp.355-368
• /
• 2016

Outliers and influential data points distort many data analysis measures. Jang and Anderson-Cook (2014) proposed a graphical method called a rework plot for exploratory analysis purpose so that there could be a possible visualization of the trace of the impact of the possible outlying and/or influential data points on the univariate/bivariate data analysis and regression. They developed 3-D plot as well as pairwise plot for the appropriate measures of interest. This paper further extends their approach to identify its strength. We can use rework plots as a graphical exploratory data analysis tool to evaluate the impact of outliers in skewness and kurtosis of univariate data.

• Proceedings of the Korean Society of Propulsion Engineers Conference
• /
• 2017.05a
• /
• pp.984-990
• /
• 2017

Mesh screen modeling and liquid propellant discharge simulation of surface tension tank were performed using commercial CFD software Flow-3d. $350{\times}2600$, $400{\times}3000$ and $510{\times}3600$ DTW mesh screen were modeled using macroscopic porous media model. Porosity, capillary pressure, and drag coefficient were assigned for each mesh screen model, and bubble point simulations were performed. The mesh screen model was validated with the experimental data. Based on the screen modeling, liquid propellant discharge simulation from PMD tank was performed. NTO was assigned as the liquid propellant, and void was set to flow into the tank inlet to achieve an initial volume flow rate of liquid propellant in $3{\times}10^{-3}g$ acceleration condition. The intial flow pressure drop through the mesh screen was approximately 270 Pa, and the pressure drop increased with time. Liquid propellant discharge was sustained until the flow pressure drop reached approximately 630 Pa, which was near the estimated bubble point value of the screen model.

• Atmosphere
• /
• v.25 no.1
• /
• pp.149-154
• /
• 2015

As an attempt to improve fog predictability at Incheon International Airport (IIA) we couple the 3D weather forecasting model currently operational in Korea Meteorological Administration (regional Unified Model, UM_RE) with a 1D turbulence model (PAFOG). The coupling is done by extracting the meteorological data from the 3D model and properly inserting them in the PAFOG model as initial conditions and external forcing. The initial conditions include surface temperature, 2 m temperature and dew point temperature, geostrophic wind at 850 hPa and vertical profiles of temperature and dew point temperature. Moisture and temperature advections are included as external forcing and updated every hr. To validate the performance of the coupled system, simulation results of the coupled system are compared to those of the 3D model alone for the 22 sea fog cases observed over the Yellow Sea. Three statistical indices, i.e., Root Mean Square Error (RMSE), linear correlation coefficient (R) and Critical Success Index (CSI), are examined, and they all indicate that the coupled system performs better than the 3D model alone. These are certainly promising results but more improvement is required before the coupled system can actually be used as an operational fog forecasting model. For the RMSE, R, and CSI values for the coupled system are still not good enough for operational fog forecast.

• International conference on construction engineering and project management
• /
• 2022.06a
• /
• pp.792-799
• /
• 2022

Building Information Modeling (BIM) technology is a key component of modern construction engineering and project management workflows. As-is BIM models that represent the spatial reality of a project site can offer crucial information to stakeholders for construction progress monitoring, error checking, and building maintenance purposes. Geometric methods for automatically converting raw scan data into BIM models (Scan-to-BIM) often fail to make use of higher-level semantic information in the data. Whereas, semantic segmentation methods only output labels at the point level without creating object level models that is necessary for BIM. To address these issues, this research proposes a hybrid semantic-geometric approach for clutter-resistant floorplan generation from laser-scanned building point clouds. The input point clouds are first pre-processed by normalizing the coordinate system and removing outliers. Then, a semantic segmentation network based on PointNet++ is used to label each point as ceiling, floor, wall, door, stair, and clutter. The clutter points are removed whereas the wall, door, and stair points are used for 2D floorplan generation. A region-growing segmentation algorithm paired with geometric reasoning rules is applied to group the points together into individual building elements. Finally, a 2-fold Random Sample Consensus (RANSAC) algorithm is applied to parameterize the building elements into 2D lines which are used to create the output floorplan. The proposed method is evaluated using the metrics of precision, recall, Intersection-over-Union (IOU), Betti error, and warping error.

• Journal of the Korean Society of Clothing and Textiles
• /
• v.32 no.2
• /
• pp.340-347
• /
• 2008

A prototype of 3D clothing design system with a direct pattern development function was suggested, reflecting intuitive design functions and design modifications while considering the fit of clothing patterns with the 3D human body in the virtual 3D space. The research method was as follows. Clothing models were created using a 3D design tool, 3ds max on the surface of 3D human body model made by scanning an actual human body. 3D illustrations were completed by revising the fit and sizing of the human body and clothing models. 2D T-shirt pattern was produced 3D illustrations using from a 3D scanning data modeling solution RapidForm 2004, a 2D conversion program for 3D data called 2C-AN, and Yuka CAD. As a result, the following conclusions were made. The fit of the clothing and human body can be adjusted by reflecting individual body figure characteristics and 3D illustrations over the actual 3D body model. Furthermore, intuitive design support functions were intensified overcoming the weak point of existing 3D clothing design system by developing the direct clothing design in the virtual 3D space. 3D illustration design modifications can be directly reflected on clothing patterns from 3D illustrations by 3D clothing design system developed in this study.

• Journal of the Korea Computer Graphics Society
• /
• v.29 no.3
• /
• pp.93-103
• /
• 2023

The problem of reconstructing three-dimensional models of people and objects from color and depth images captured by low-cost RGB-D cameras has long been an active research area in computer graphics. Color and depth images captured by low-cost RGB-D cameras are represented as point clouds in three-dimensional space, which correspond to discrete values in a continuous three-dimensional space and require additional surface reconstruction compared to rendering using polygonal models. In this paper, we propose an effective ray-tracing based technique for visualizing point clouds rather than polygonal models. In particular, our method shows the possibility of an effective rendering method even in mobile environment which has limited performance due to processor heat and lack of battery.

• Journal of the Korean Society of Manufacturing Technology Engineers
• /
• v.26 no.1
• /
• pp.108-113
• /
• 2017

In this paper, we proposed a three-dimensional(3D) scanning system based on laser vision technique for 3D model reconstruction. The proposed scanning system consists of line laser, camera, and turntable. We implemented the 3D scanning system using low quality elements. Although these are low quality elements, we reduced the 3D data reconstruction errors greatly using two methods. First, we developed a maximum brightness detection algorithm. This algorithm extracts the maximum brightness of the line laser to obtain the shape of the object. Second, we designed a new laser control device. This device helps to adjust the relative position of the turntable and line laser. These two methods greatly reduce the measuring noise. As a result, point cloud data can be obtained without complicated calculations.

• Journal of Sasang Constitutional Medicine
• /
• v.20 no.1
• /
• pp.42-47
• /
• 2008

1. Objectives The Body Shape and Feature is one of the important standard for classification of Sasang Constitutions. In order to evaluate one's Body Shape and Feature objectively we have been developing the Body Measuring Machine. Now we develop the 3D Automatic Body Measuring Machine(3D-ABMM). So we make an analysis of the 3D-ABMM's Accuracy. 2. Methods By using the 3D-ABMM and Vivid 9i(3D laser scanner, Konica Minolta) we have a surface scan of the three objects which are the upper body of the female and male Manikin and a male model. We overlap each scan data using the RapidForm2006 (3D scan data solution, INUS Technology) and calculate the average distance and standard deviation between the same point of each scan data. 3. Results and Conclusions In the female Manikin, the average distance is 0.84mm and the standard deviation is 1.16mm and the maximum distance is 10.68mm. In the male Manikin, the average distance is 1.12mm and the standard deviation is 1.19mm and the maximum distance is 12.00mm. In the male model, the average distance is 3.26mm and the standard deviation is 2.59mm and the maximum distance is 12.75mm. From the results, 3D-ABMM has good accuracy for scanning body and will be a usable hardware of the 3D Automatic Body Analysis Machine.