• Journal of the Korea Institute of Information Security & Cryptology
• /
• v.33 no.6
• /
• pp.1099-1110
• /
• 2023

The recognition system of autonomous driving and robot navigation performs vision work such as object recognition, tracking, and lane detection after multi-sensor fusion to improve performance. Currently, research on a deep learning model based on the fusion of a camera and a lidar sensor is being actively conducted. However, deep learning models are vulnerable to adversarial attacks through modulation of input data. Attacks on the existing multi-sensor-based autonomous driving recognition system are focused on inducing obstacle detection by lowering the confidence score of the object recognition model.However, there is a limitation that an attack is possible only in the target model. In the case of attacks on the sensor fusion stage, errors in vision work after fusion can be cascaded, and this risk needs to be considered. In addition, an attack on LIDAR's point cloud data, which is difficult to judge visually, makes it difficult to determine whether it is an attack. In this study, image scaling-based camera-lidar We propose an attack method that reduces the accuracy of LCCNet, a fusion model (camera-LiDAR calibration model). The proposed method is to perform a scaling attack on the point of the input lidar. As a result of conducting an attack performance experiment by size with a scaling algorithm, an average of more than 77% of fusion errors were caused.

• Journal of Intelligence and Information Systems
• /
• v.29 no.1
• /
• pp.107-119
• /
• 2023

The technology that recognizes a soldier's motion and movement status has recently attracted large attention as a combination of wearable technology and artificial intelligence, which is expected to upend the paradigm of troop management. The accuracy of state determination should be maintained at a high-end level to make sure of the expected vital functions both in a training situation; an evaluation and solution provision for each individual's motion, and in a combat situation; overall enhancement in managing troops. However, when input data is given as a timer series or sequence, existing feedforward networks would show overt limitations in maximizing classification performance. Since human behavior data (3-axis accelerations and 3-axis angular velocities) handled for military motion recognition requires the process of analyzing its time-dependent characteristics, this study proposes a high-performance data-driven classifier which utilizes the long-short term memory to identify the order dependence of acquired data, learning to classify eight representative military operations (Sitting, Standing, Walking, Running, Ascending, Descending, Low Crawl, and High Crawl). Since the accuracy is highly dependent on a network's learning conditions and variables, manual adjustment may neither be cost-effective nor guarantee optimal results during learning. Therefore, in this study, we optimized hyperparameters using Bayesian optimization for maximized generalization performance. As a result, the final architecture could reduce the error rate by 62.56% compared to the existing network with a similar number of learnable parameters, with the final accuracy of 98.39% for various military operations.

• The Journal of Korea Institute of Information, Electronics, and Communication Technology
• /
• v.16 no.5
• /
• pp.331-337
• /
• 2023

In this paper, we propose a deep learning model to detect lumbar 3 (L3) CT images to determine the occurrence and degree of sarcopenia. In addition, we would like to propose an optimization technique that uses oversampling ratio and class weight as design parameters to address the problem of performance degradation due to data imbalance between L3 level and non-L3 level portions of CT data. In order to train and test the model, a total of 150 whole-body CT images of 104 prostate cancer patients and 46 bladder cancer patients who visited Gangneung Asan Medical Center were used. The deep learning model used ResNet50, and the design parameters of the optimization technique were selected as six types of model hyperparameters, data augmentation ratio, and class weight. It was confirmed that the proposed optimization-based L3 level extraction model reduced the median L3 error by about 1.0 slices compared to the control model (a model that optimized only 5 types of hyperparameters). Through the results of this study, accurate L3 slice detection was possible, and additionally, we were able to present the possibility of effectively solving the data imbalance problem through oversampling through data augmentation and class weight adjustment.

• Journal of Internet of Things and Convergence
• /
• v.10 no.4
• /
• pp.101-108
• /
• 2024

Indoor localization is a critical component for numerous applications, ranging from navigation in large buildings to emergency response. This paper presents an enhanced Pedestrian Dead Reckoning (PDR) scheme using smartphone sensors, integrating neural network-aided motion recognition, Kalman filter-based error correction, and multi-sensor data fusion. The proposed system leverages data from the accelerometer, magnetometer, gyroscope, and barometer to accurately estimate a user's position and orientation. A neural network processes sensor data to classify motion modes and provide real-time adjustments to stride length and heading calculations. The Kalman filter further refines these estimates, reducing cumulative errors and drift. Experimental results, collected using a smartphone across various floors of University, demonstrate the scheme's ability to accurately track vertical movements and changes in heading direction. Comparative analyses show that the proposed CNN-LSTM model outperforms conventional CNN and Deep CNN models in angle prediction. Additionally, the integration of barometric pressure data enables precise floor level detection, enhancing the system's robustness in multi-story environments. Proposed comprehensive approach significantly improves the accuracy and reliability of indoor localization, making it viable for real-world applications.

• Korean Journal of Remote Sensing
• /
• v.37 no.3
• /
• pp.603-614
• /
• 2021

Recently, the surface temperature in the seas around Korea has been continuously rising. This temperature rise causes changes in fishery resources and affects leisure activities such as fishing. In particular, high temperatures lead to the occurrence of red tides, causing severe damage to ocean industries such as aquaculture. Meanwhile, changes in sea temperature are closely related to military operation to detect submarines. This is because the degree of diffraction, refraction, or reflection of sound waves used to detect submarines varies depending on the ocean mixed layer. Currently, research on the prediction of changes in sea water temperature is being actively conducted. However, existing research is focused on predicting only the surface temperature of the ocean, so it is difficult to identify fishery resources according to depth and apply them to military operations such as submarine detection. Therefore, in this study, we predicted the temperature of the ocean mixed layer at a depth of 38m by using temperature data for each water depth in the upper mixed layer and meteorological data such as temperature, atmospheric pressure, and sunlight that are related to the surface temperature. The data used are meteorological data and sea temperature data by water depth observed from 2016 to 2020 at the IEODO Ocean Research Station. In order to increase the accuracy and efficiency of prediction, LSTM (Long Short-Term Memory), which is known to be suitable for time series data among deep learning techniques, was used. As a result of the experiment, in the daily prediction, the RMSE (Root Mean Square Error) of the model using temperature, atmospheric pressure, and sunlight data together was 0.473. On the other hand, the RMSE of the model using only the surface temperature was 0.631. These results confirm that the model using meteorological data together shows better performance in predicting the temperature of the upper ocean mixed layer.

• Journal of Veterinary Clinics
• /
• v.33 no.2
• /
• pp.107-112
• /
• 2016

The objective of this study was to determine potential contributing factors associated with biosecurity level of farrow-to-finish pig farms and to develop a classification tree model to explore how these factors related to each other based on prediction model. To this end, the author analyzed data (n = 193) extracted from a cross-sectional study of 344 farrow-to-finish farms which was conducted between March and September 2014 aimed to explore swine disease status at farm level. Standardized questionnaires with information about basic demographical data and management practices were collected in each farm by on-site visit of trained veterinarians. For the classification of the data sets regarding biosecurity level as a dependent variable and predictor variables, Chi-squared Automatic Interaction Detection (CHAID) algorithm was applied for modeling classification tree. The statistics of misclassification risk was used to evaluate the fitness of the model in terms of prediction results. Categorical multivariate input data (40 variables) was used to construct a classification tree, and the target variable was biosecurity level dichotomized into low versus high. In general, the level of biosecurity was lower in the majority of farms studied, mainly due to the limited implementation of on-farm basic biosecurity measures aimed at controlling the potential introduction and transmission of swine diseases. The CHAID model illustrated the relative importance of significant predictors in explaining the level of biosecurity; maintenance of medical records of treatment and vaccination, use of dedicated clothing to enter the farm, installing fence surrounding the farm perimeter, and periodic monitoring of the herd using written biosecurity plan in place. The misclassification risk estimate of the prediction model was 0.145 with the standard error of 0.025, indicating that 85.5% of the cases could be classified correctly by using the decision rule based on the current tree. Although CHAID approach could provide detailed information and insight about interactions among factors associated with biosecurity level, further evaluation of potential bias intervened in the course of data collection should be included in future studies. In addition, there is still need to validate findings through the external dataset with larger sample size to improve the external validity of the current model.

• The Journal of The Korea Institute of Intelligent Transport Systems
• /
• v.22 no.5
• /
• pp.1-18
• /
• 2023

With the recent developments in big data and deep learning, a variety of traffic information is collected widely and used for traffic operations. In particular, long short-term memory (LSTM) is used in the field of traffic information prediction with time series characteristics. Since trends, seasons, and cycles differ due to the nature of time series data input for an LSTM, a trial-and-error method based on characteristics of the data is essential for prediction models based on time series data in order to find hyperparameters. If a methodology is established to find suitable hyperparameters, it is possible to reduce the time spent in constructing high-accuracy models. Therefore, in this study, a traffic information prediction model is developed based on highway vehicle detection system (VDS) data and LSTM, and an impact assessment is conducted through changes in the LSTM evaluation indicators for each hyperparameter. In addition, a methodology for finding hyperparameters suitable for predicting highway traffic information in the transportation field is presented.

• Journal of the Korean earth science society
• /
• v.37 no.7
• /
• pp.420-433
• /
• 2016

The purpose of this study is to improve the calibration matrixes of 2-D and 3-D convective rainfall rates (CRR) using the brightness temperature of the infrared $10.8{\mu}m$ channel (IR), the difference of brightness temperatures between infrared $10.8{\mu}m$ and vapor $6.7{\mu}m$ channels (IR-WV), and the normalized reflectance of the visible channel (VIS) from the COMS satellite and rainfall rate from the weather radar for the period of 75 rainy days from April 22, 2011 to October 22, 2011 in Korea. Especially, the rainfall rate data of the weather radar are used to validate the new 2-D and 3-DCRR calibration matrixes suitable for the Korean peninsula for the period of 24 rainy days in 2011. The 2D and 3D calibration matrixes provide the basic and maximum CRR values ($mm\;h^{-1}$) by multiplying the rain probability matrix, which is calculated by using the number of rainy and no-rainy pixels with associated 2-D (IR, IR-WV) and 3-D (IR, IR-WV, VIS) matrixes, by the mean and maximum rainfall rate matrixes, respectively, which is calculated by dividing the accumulated rainfall rate by the number of rainy pixels and by the product of the maximum rain rate for the calibration period by the number of rain occurrences. Finally, new 2-D and 3-D CRR calibration matrixes are obtained experimentally from the regression analysis of both basic and maximum rainfall rate matrixes. As a result, an area of rainfall rate more than 10 mm/h is magnified in the new ones as well as CRR is shown in lower class ranges in matrixes between IR brightness temperature and IR-WV brightness temperature difference than the existing ones. Accuracy and categorical statistics are computed for the data of CRR events occurred during the given period. The mean error (ME), mean absolute error (MAE), and root mean squire error (RMSE) in new 2-D and 3-D CRR calibrations led to smaller than in the existing ones, where false alarm ratio had decreased, probability of detection had increased a bit, and critical success index scores had improved. To take into account the strong rainfall rate in the weather events such as thunderstorms and typhoon, a moisture correction factor is corrected. This factor is defined as the product of the total precipitable waterby the relative humidity (PW RH), a mean value between surface and 500 hPa level, obtained from a numerical model or the COMS retrieval data. In this study, when the IR cloud top brightness temperature is lower than 210 K and the relative humidity is greater than 40%, the moisture correction factor is empirically scaled from 1.0 to 2.0 basing on PW RH values. Consequently, in applying to this factor in new 2D and 2D CRR calibrations, the ME, MAE, and RMSE are smaller than the new ones.

• Journal of the Institute of Electronics and Information Engineers
• /
• v.49 no.9
• /
• pp.347-352
• /
• 2012

In this paper, an 2D spatial-map construction for workers identification and avoidance of AGV using the detection scheme of the spatial coordinates based on stereo camera is proposed. In the proposed system, face area of a moving person is detected from a left image among the stereo image pairs by using the YCbCr color model and its center coordinates are computed by using the centroid method and then using these data, the stereo camera embedded on the mobile robot can be controlled for tracking the moving target in real-time. Moreover, using the disparity map obtained from the left and right images captured by the tracking-controlled stereo camera system and the perspective transformation between a 3-D scene and an image plane, depth map can be detected. From some experiments on AGV driving with 240 frames of the stereo images, it is analyzed that error ratio between the calculated and measured values of the worker's width is found to be very low value of 2.19% and 1.52% on average.

• The Transactions of The Korean Institute of Electrical Engineers
• /
• v.63 no.3
• /
• pp.409-413
• /
• 2014

GIS partial discharge that occurred in the UHF band signal is effectively detected by the method to IEC60270 5pc the apparent minimum discharge can be detected over the GIS arrangement of the sensor interior and exterior of the UHF in accordance with the optimized position signal by considering the damping ratio is selected so that the signals can be obtained to be mounted. 362kV, 800kV GIS is installed on the internal and external sensors are UHF band signal attenuation is set by measuring the reference value, but the operation, 170kV case 362kV, 800kV on the basis of the measurement data and to be installed and operated. When 170kV per 1Bay by installing the built-in sensor 1 for detecting a partial discharge signal, But, GIS signal attenuation is large in the case of an internal partial discharge signal is not detected in some cases. Where the attenuation is great UHF signal of the sensor by increasing the quantity of partial discharge signals were acquired to allow relocation. The greater the spacing between the sensor and the sensor is applied simplifies the installation and reduces the cost in terms of maintenance of appropriate optimal position is calculated to detect the partial discharge signal is needed. Thus 170kV GIS signal power attenuation of a partial discharge by measuring the UHF sensor, and by relocating the proper position is calculated in accordance with the sensor signal decay rate and minimize the error of omission in detecting a partial discharge signal was optimized.