• Korean Journal of Remote Sensing
• /
• v.19 no.4
• /
• pp.307-317
• /
• 2003

This paper presents an algorithm that automatically extracts buildings among many different features on the earth surface by fusing LIDAR data with panchromatic aerial images. The proposed algorithm consists of three stages such as point level process, polygon level process, parameter space level process. At the first stage, we eliminate gross errors and apply a local maxima filter to detect building candidate points from the raw laser scanning data. After then, a grouping procedure is performed for segmenting raw LIDAR data and the segmented LIDAR data is polygonized by the encasing polygon algorithm developed in the research. At the second stage, we eliminate non-building polygons using several constraints such as area and circularity. At the last stage, all the polygons generated at the second stage are projected onto the aerial stereo images through collinearity condition equations. Finally, we fuse the projected encasing polygons with edges detected by image processing for refining the building segments. The experimental results showed that the RMSEs of building corners in X, Y and Z were 8.1cm, 24.7cm, 35.9cm, respectively.

• Journal of Internet Computing and Services
• /
• v.22 no.2
• /
• pp.59-68
• /
• 2021

As artificial intelligence(AI) technologies, which have made rapid growth recently, began to be applied to the marine environment such as ships, there have been active researches on the application of CNN-based models specialized for digital videos. In E-Navigation service, which is combined with various technologies to detect floating objects of clash risk to reduce human errors and prevent fires inside ships, real-time processing is of huge importance. More functions added, however, mean a need for high-performance processes, which raises prices and poses a cost burden on shipowners. This study thus set out to propose a method capable of processing information at a high rate while maintaining the accuracy by applying Quantization techniques of a deep learning model. First, videos were pre-processed fit for the detection of floating matters in the sea to ensure the efficient transmission of video data to the deep learning entry. Secondly, the quantization technique, one of lightweight techniques for a deep learning model, was applied to reduce the usage rate of memory and increase the processing speed. Finally, the proposed deep learning model to which video pre-processing and quantization were applied was applied to various embedded boards to measure its accuracy and processing speed and test its performance. The proposed method was able to reduce the usage of memory capacity four times and improve the processing speed about four to five times while maintaining the old accuracy of recognition.

• Clean Technology
• /
• v.25 no.2
• /
• pp.140-146
• /
• 2019

Flammable substances, such as organic solvents, are commonly used in laboratories and industrial processes. The flash point of flammable liquid mixtures is a very important parameter for characterizing the ignition and explosion hazards, and the flash points of mixtures of $C_2{\sim}C_3$ alcohols and 2,2,4-trimethylpentane were measured in the present study. The 2,2,4-trimethylpentane is an important component of gasoline and is frequently used in the petroleum industry as a solvent. Lower flash point data were measured for the binary systems {ethanol + 2,2,4-trimethylpentane}, {1-propanol + 2,2,4-trimethylpentane}, and {2-propanol + 2,2,4-trimethylpentane}. The flash point measurements were carried out according to the standard test method (ASTM D3278) using a Stanhope-Seta closed cup flash point tester. The measured flash points were compared with the predicted values calculated using Raoult's law and also following $G^E$ models: Wilson, Non-Random Two Liquid (NRTL) and UNIversal QUAsiChemical (UNIQUAC). These models were able to predict the experimental flash points for different compositions of {$C_2{\sim}C_3$ alcohols + 2,2,4-trimethylpentane} mixtures with minimal deviations. The average absolute deviation between the predicted and measured lower flash point was less than 1.28 K. A minimum flash point behaviour was observed in all of the systems as in the many observed cases for the hydrocarbon and alcohol mixtures.

• Korean Chemical Engineering Research
• /
• v.57 no.2
• /
• pp.274-282
• /
• 2019

In recent years, big data analysis has been expanded to include automatic control through reinforcement learning as well as prediction through modeling. Research on the utilization of image data is actively carried out in various industrial fields such as chemical, manufacturing, agriculture, and bio-industry. In this paper, we applied NASNet, which is an AutoML reinforced learning algorithm, to DeepU-Net neural network that modified U-Net to improve image semantic segmentation performance. We used BRATS2015 MRI data for performance verification. Simulation results show that DeepU-Net has more performance than the U-Net neural network. In order to improve the image segmentation performance, remove dropouts that are typically applied to neural networks, when the number of kernels and filters obtained through reinforcement learning in DeepU-Net was selected as a hyperparameter of neural network. The results show that the training accuracy is 0.5% and the verification accuracy is 0.3% better than DeepU-Net. The results of this study can be applied to various fields such as MRI brain imaging diagnosis, thermal imaging camera abnormality diagnosis, Nondestructive inspection diagnosis, chemical leakage monitoring, and monitoring forest fire through CCTV.

• KIPS Transactions on Software and Data Engineering
• /
• v.7 no.12
• /
• pp.485-496
• /
• 2018

So far, the deep learning, a field of artificial intelligence, has achieved remarkable results in solving problems from unstructured data. However, it is difficult to comprehensively judge situations like humans, and did not reach the level of intelligence that deduced their relations and predicted the next situation. Recently, deep neural networks show that artificial intelligence can possess powerful relational reasoning that is core intellectual ability of human being. In this paper, to analyze and observe the performance of Relation Networks (RN) among the neural networks for relational reasoning, two types of RN-based deep neural network models were constructed and compared with the baseline model. One is a visual question answering RN model using Sort-of-CLEVR and the other is a text-based question answering RN model using bAbI task. In order to maximize the performance of the RN-based model, various performance improvement experiments such as hyper parameters tuning have been proposed and performed. The effectiveness of the proposed performance improvement methods has been verified by applying to the visual QA RN model and the text-based QA RN model, and the new domain model using the dialogue-based LL dataset. As a result of the various experiments, it is found that the initial learning rate is a key factor in determining the performance of the model in both types of RN models. We have observed that the optimal initial learning rate setting found by the proposed random search method can improve the performance of the model up to 99.8%.

• Journal of the Korean Society of Marine Environment & Safety
• /
• v.24 no.7
• /
• pp.953-959
• /
• 2018

This paper deals with controlling surge oscillations of a mooring vessel system under large external disturbances such as wind, waves and currents. A control synthesis based on Sliding Mode Control (SMC) with a Super-Twisting Algorithm (STA) has been applied to suppress nonlinear surge oscillations of a two-point mooring system. Despite the advantages of robustness against parameter uncertainties and disturbances for SMC, chattering is the main drawback for implementing sliding mode controllers. First-order SMC shows convergence within the desired level of accuracy, in which chattering is the main obstacle related to the destructive phenomenon. Alternatively, STA completely eliminates chattering phenomenon with high accuracy even for large disturbances. SMC based on STA is an effective tool for the motion control of a nonlinear mooring system because it avoids the chattering problems of a first-order sliding mode controller. In addition, the error trajectories of controlled mooring systems implemented by means of STA form in the bounded region. Finally, the control gain effect of STA can be observed in sliding surface and position trajectory errors.

• Journal of Broadcast Engineering
• /
• v.24 no.3
• /
• pp.463-471
• /
• 2019

This paper proposes a method for reducing the complexity of LIC (Local Illuminance Compensation) for bi-directional inter prediction. The LIC performs local illumination compensation using neighboring reconstruction samples of the current block and the reference block to improve the accuracy of the inter prediction. Since the weight and offset required for local illumination compensation are calculated at both sides of the encoder and decoder using the reconstructed samples, there is an advantage that the coding efficiency is improved without signaling any information. Since the weight and the offset are obtained in the encoding prediction step and the decoding step, encoder and decoder complexity are increased. This paper proposes two methods for low complexity LIC. The first method is a method of applying illumination compensation with offset only in bi-directional prediction, and the second is a method of applying LIC after weighted average step of reference block obtained by bidirectional prediction. To evaluate the performance of the proposed method, BD-rate is compared with BMS-2.0.1 using B, C, and D classes of MPEG standard experimental image under RA (Random Access) condition. Experimental results show that the proposed method reduces the average of 0.29%, 0.23%, 0.04% for Y, U, and V in terms of BD-rate performance compared to BMS-2.0.1 and encoding/decoding time is almost same. Although the BD-rate was lost, the calculation complexity of the LIC was greatly reduced as the multiplication operation was removed and the addition operation was halved in the LIC parameter derivation process.

• Journal of the Korea Institute of Information Security & Cryptology
• /
• v.21 no.1
• /
• pp.135-141
• /
• 2011

Recent advances in mobile devices and development of various mobile applications dealing with private information of users made user authentication in mobile devices a very important issue. This paper presents a new user authentication method based on touch screen interfaces. This method uses for authentication the PIN digits as well as the exact locations the user touches to input these digits. Our method is fully compatible with the regular PIN entry method which uses numeric keypads, and it provides better usability than the behavioral biometric schemes because its PIN registration process is much simpler. According to our experiments, our method guarantees EERs of 12.8%, 8.3%, and 9.3% for 4-digit PINs, 6-digit PINs, and 11-digit cell phone numbers, respectively, under the extremely conservative assumption that all users have the same PIN digits and cell phone numbers. Thus we can guarantee much higher performance in identification functionality by applying this result to a more practical situation where every user uses distinct PIN and sell phone number. Finally, our method is far more secure than the regular PIN entry method, which is verified by our experiments where attackers are required to recover a PIN after observing the PIN entry processes of the regular PIN and our method under the same level of security parameters.

• Journal of the Computational Structural Engineering Institute of Korea
• /
• v.34 no.4
• /
• pp.175-182
• /
• 2021

There have been many instances of damage to buildings with soft stories, and it is important to consider vertically irregular buildings when evaluating the seismic performance of existing buildings. However, because conventional methods do not easily reflect vertical irregularities with sufficient accuracy, it is possible to underestimate or overestimate the seismic performance of buildings with vertical irregularities. This study aims to develop a seismic performance evaluation method for vertically irregular buildings using the stiffness-based soft story ratio (SSR), which is a parameter that represents the ratio of the demand and the capacity for displacement and refers to the ratio of displacement concentration in buildings. The seismic performance evaluation method developed in this study is compared with the conventional seismic performance evaluation method for four piloti buildings, using the first-floor column as a variable. Conventional seismic performance evaluation methods often overestimate the seismic performance for models in which vertical irregularities are maximized. However, results of the proposed seismic performance evaluation method are identical to those from a detailed evaluation for all models. Therefore, it is considered that the proposed seismic performance evaluation method can provide more precise seismic performance evaluation results than conventional methods in the case of piloti buildings, where vertical irregularities are maximized.

• Journal of Convergence for Information Technology
• /
• v.11 no.5
• /
• pp.30-37
• /
• 2021

Various devices are connected to the Internet, and attacks using the Internet are occurring. Among such attacks, there are attacks that use malicious URLs to make users access to wrong phishing sites or distribute malicious viruses. Therefore, how to detect such malicious URL attacks is one of the important security issues. Among recent deep learning technologies, neural networks are showing good performance in image recognition, speech recognition, and pattern recognition. This neural network can be applied to research that analyzes and detects patterns of malicious URL characteristics. In this paper, performance analysis according to various parameters was performed on a method of detecting malicious URLs using neural networks. In this paper, malicious URL detection performance was analyzed while changing the activation function, learning rate, and neural network structure. The experimental data was crawled by Alexa top 1 million and Whois to build the data, and the machine learning library used TensorFlow. As a result of the experiment, when the number of layers is 4, the learning rate is 0.005, and the number of nodes in each layer is 100, the accuracy of 97.8% and the f1 score of 92.94% are obtained.