• KIPS Transactions on Software and Data Engineering
• /
• v.12 no.12
• /
• pp.519-524
• /
• 2023

In hashing-based image retrieval, the hash code of a manipulated image is different from the original image, making it difficult to search for the same image. This paper proposes and evaluates a self-supervised deephashing model that generates perceptual hash codes from feature information such as texture, shape, and color of images. The comparison models are autoencoder-based variational inference models, but the encoder is designed with a fully connected layer, convolutional neural network, and transformer modules. The proposed model is a variational inference model that includes a SimAM module of extracting geometric patterns and positional relationships within images. The SimAM module can learn latent vectors highlighting objects or local regions through an energy function using the activation values of neurons and surrounding neurons. The proposed method is a representation learning model that can generate low-dimensional latent vectors from high-dimensional input images, and the latent vectors are binarized into distinguishable hash code. From the experimental results on public datasets such as CIFAR-10, ImageNet, and NUS-WIDE, the proposed model is superior to the comparative model and analyzed to have equivalent performance to the supervised learning-based deephashing model. The proposed model can be used in application systems that require low-dimensional representation of images, such as image search or copyright image determination.

• The Journal of the Acoustical Society of Korea
• /
• v.25 no.7
• /
• pp.346-355
• /
• 2006

Acoustic data from a shallow water experiment in the East Sea of Korea (MAPLE IV) is Processed to investigate the Performance of matched-field geo-acoustic inversion and source localization. The receiver array consists of two legs as in an L-shape. one vertical and the other horizontal lying on the seabed. Narrowband multi-tone CW source was towed along a slightly inclined bathymetry track. The matched-field geo-acoustic inversion includes comparisons between three processing techniques. all based on the Bartlett processor as; (1) the coherent processing of the data from the full array, (2) the incoherent Product of each output from both the horizontal and vertical arrays, and (3) the cross correlation between the horizontal and vertical arrays. as well as processing each array leg separately. To verify the inversion results. matched-field source localization for low level source signal components were performed using the same Processors used at the inversion stage.

• The Journal of the Acoustical Society of Korea
• /
• v.29 no.7
• /
• pp.458-467
• /
• 2010

In this paper, a new audio reproduction system was developed in which the cross-talk signals would be reasonably cancelled at an arbitrary listener position. To adaptively remove the cross-talk signals according to the listener's position, a method of tracking the listener position was employed. This was achieved using the two microphones, where the listener direction was estimated using the time-delay between the two signals from the two microphones, respectively. Moreover, room reverberation effects were taken into consideration where linear prediction analysis was involved. To remove the cross-talk signals at the left-and right-ears, the paths between the sources and the ears were represented using the KEMAR head-related transfer functions (HRTFs) which were measured from the artificial dummy head. To evaluate the usefulness of the proposed listener tracking system, the performance of cross-talk cancellation was evaluated at the estimated listener positions. The performance was evaluated in terms of the channel separation ration (CSR), a -10 dB of CSR was experimentally achieved although the listener positions were more or less deviated. A real-time system was implemented using a floating-point digital signal processor (DSP). It was confirmed that the average errors of the listener direction was 5 degree and the subjects indicated that 80 % of the stimuli was perceived as the correct directions.

• Information Systems Review
• /
• v.25 no.1
• /
• pp.93-110
• /
• 2023

Recently, attempts have been made to convert unstructured text into vectors and to analyze vast amounts of natural language for various purposes. In particular, the demand for analyzing texts in specialized domains is rapidly increasing. Therefore, studies are being conducted to analyze specialized and general-purpose documents simultaneously. To analyze specific terms with general terms, it is necessary to align the embedding space of the specific terms with the embedding space of the general terms. So far, attempts have been made to align the embedding of specific terms into the embedding space of general terms through a transformation matrix or mapping function. However, the linear transformation based on the transformation matrix showed a limitation in that it only works well in a local range. To overcome this limitation, various types of nonlinear vector alignment methods have been recently proposed. We propose a vector alignment model that matches the embedding space of specific terms to the embedding space of general terms through end-to-end learning that simultaneously learns the autoencoder and regression model. As a result of experiments with R&D documents in the "Healthcare" field, we confirmed the proposed methodology showed superior performance in terms of accuracy compared to the traditional model.

• The Journal of the Acoustical Society of Korea
• /
• v.43 no.2
• /
• pp.214-224
• /
• 2024

Sonobuoys are disposable devices that utilize sound waves for information gathering, detecting engine noises, and capturing various acoustic characteristics. They play a crucial role in accurately detecting underwater targets, making them effective detection systems in anti-submarine warfare. Existing sonobuoy deployment methods in multistatic systems often rely on fixed patterns or heuristic-based rules, lacking efficiency in terms of the number of sonobuoys deployed and operational time due to the unpredictable mobility of the underwater targets. Thus, this paper proposes an optimal sonobuoy placement strategy for Unmanned Aerial Vehicles (UAVs) to overcome the limitations of conventional sonobuoy deployment methods. The proposed approach utilizes reinforcement learning in a simulation-based experimental environment that considers the movements of the underwater targets. The Unity ML-Agents framework is employed, and the Proximal Policy Optimization (PPO) algorithm is utilized for UAV learning in a virtual operational environment with real-time interactions. The reward function is designed to consider the number of sonobuoys deployed and the cost associated with sound sources and receivers, enabling effective learning. The proposed reinforcement learning-based deployment strategy compared to the conventional sonobuoy deployment methods in the same experimental environment demonstrates superior performance in terms of detection success rate, deployed sonobuoy count, and operational time.

• The Journal of the Convergence on Culture Technology
• /
• v.10 no.1
• /
• pp.585-590
• /
• 2024

In this study, rail surface defects are increasing due to the aging of urban railway rails, but in the detailed guidelines for track performance evaluation established by the country, rail surface damage is inspected with the naked eye of engineers and simple measuring tools. With the recent enactment of the Track Diagnosis Act, a large budget has been invested and the volume of rail diagnosis is rapidly increasing, but it is difficult to secure the reliability of diagnosis results using labor-intensive visual inspection techniques. It is very important to discover defects in the rail surface through periodic track tours and visual inspection. However, evaluating the severity of defects on the rail surface based on the subjective judgment of the inspector has significant limitations in predicting damage inside the rail. In this study, the rail internal crack characteristics due to rail surface damage were studied. In field measurements, rail surface damage locations were selected, samples of various damage types were collected, and the rail surface damage status was evaluated. In indoor testing, we intend to analyze the correlation between rail surface defects and internal defects using a electron scanning microscope (SEM). To determine the crack growth rate of urban railway rails currently in use, the Gaussian probability density function was applied and analyzed.

• The Journal of the Acoustical Society of Korea
• /
• v.43 no.3
• /
• pp.314-323
• /
• 2024

Recently, a score-penalty method has been used for the acoustic passive tracking of marine mammals. The interesting aspect of this technique lies in the loss function, which has a penalty term representing the mismatch between the measured signal and the modeled signal, while the traditional time-domain matched-field processing is positively considering the match between them. In this study, we apply the score-penalty method into the depth estimation of a passive target with a known source waveform. Assuming deep ocean environments with uncertainties in the sound speed profile, we evaluate the score-penalty method, comparing it with the time-domain matched field processing method. We shows that the score-penalty method is more accurate than the time-domain matched field processing method in the ocean environment with weak mismatch of sound speed profile, and has better efficiency. However, in the ocean enviroment with strong mismatch of the sound speed profile, the score-penalty method also fails in the depth estimation of a target, similar to the time-domain matched-field processing method.

• Journal of Bio-Environment Control
• /
• v.33 no.1
• /
• pp.1-11
• /
• 2024

In this study, five artificial intelligence (AI) models: Inception v3, SqueezeNet (local), VGG-16, Painters, and DeepLoc were used to classify tea leaf diseases. Eight image categories were used: healthy, algal leaf spot, anthracnose, bird's eye spot, brown blight, gray blight, red leaf spot, and white spot. Software used in this study was Orange 3 which functions as a Python library for visual programming, that operates through an interface that generates workflows to visually manipulate and analyze the data. The precision of each AI model was recorded to select the ideal AI model. All models were trained using the Adam solver, rectified linear unit activation function, 100 neurons in the hidden layers, 200 maximum number of iterations in the neural network, and 0.0001 regularizations. To extend the functionality of Orange 3, new add-ons can be installed and, this study image analytics add-on was newly added which is required for image analysis. For the training model, the import image, image embedding, neural network, test and score, and confusion matrix widgets were used, whereas the import images, image embedding, predictions, and image viewer widgets were used for the prediction. Precisions of the neural networks of the five AI models (Inception v3, SqueezeNet (local), VGG-16, Painters, and DeepLoc) were 0.807, 0.901, 0.780, 0.800, and 0.771, respectively. Finally, the SqueezeNet (local) model was selected as the optimal AI model for the detection of tea diseases using tea leaf images owing to its high precision and good performance throughout the confusion matrix.

• Journal of the Korea institute for structural maintenance and inspection
• /
• v.28 no.5
• /
• pp.47-53
• /
• 2024

This study investigates the noise reduction effects of various road pavement methods to mitigate traffic noise caused by the increasing proximity between roads and residential areas in urban environments. The noise characteristics of four types of road pavement-Dense Asphalt Concrete (DAC), Double Layer Porous Asphalt Concrete (DLPAC), Transverse Tining Concrete (TTC), and Exposed Aggregate Concrete (EAC)-were evaluated using CPX close-proximity noise and pass-by noise measurements. The CPX measurements showed that noise levels increased logarithmically with vehicle speed for all pavements. Specifically, DLPAC demonstrated higher noise levels in the low-frequency range below 800 Hz and lower noise levels in the high-frequency range, which is attributed to resonance effects within the internal pores of the pavement and the reduction of compression and expansion noise. In pass-by noise measurements, DLPAC exhibited higher low-frequency noise compared to DAC, likely due to pavement durability deterioration and the influence of external environmental noise. The results indicate that the CPX measurement method is more effective in evaluating road noise performance as it better reflects the impact of vehicle speed. However, since the study was conducted under limited site conditions, further research across various sites and conditions is necessary to enhance reliability.

• Progress in Medical Physics
• /
• v.20 no.4
• /
• pp.191-198
• /
• 2009

In this study, we evaluated an edge detector for small-beam dosimetry. We measured the dose linearity, dose rate dependence, output factor, beam profiles, and percentage depth dose using an edge detector (Model 1118 Edge) for 6-MV photon beams at different field sizes and depths. The obtained values were compared with those obtained using a standard volume ionization chamber (CC13) and photon diode detector (PFD). The dose linearity results for the three detectors showed good agreement within 1%. The edge detector had the best linearity of ${\pm}0.08%$. The edge detector and PFD showed little dose rate dependency throughout the range of 100~600 MU/min, while CC13 showed a significant discrepancy of approximately -5% at 100 MU/min. The output factors of the three detectors showed good agreement within 1% for the tested field sizes. However, the output factor of CC13 compared to the other two detectors had a maximum difference of 21% for small field sizes (${\sim}4{\times}4\;cm^2$). When analyzing the 20~80% penumbra, the penumbra measured using CC13 was approximately two times wider than that using the edge detector for all field sizes. The width measured using PFD was approximately 30% wider for all field sizes. Compared to the edge detector, the 10~90% penumbras measured using the CC13 and PFD were approximately 55% and 19% wider, respectively. The full width at half maximum (FWHM) of the edge detector was close to the real field size, while the other two detectors measured values that were 8~10% greater for all field sizes. Percentage depth doses measured by the three detectors corresponded to each other for small beams. Based on the results, we consider the edge detector as an appropriate small-beam detector, while CC13 and PFD can lead to some errors when used for small beam fields under $4{\times}4\;cm^2$.