• ETRI Journal
• /
• v.44 no.6
• /
• pp.1034-1046
• /
• 2022

The learning-based multiview stereo (MVS) methods for three-dimensional (3D) reconstruction generally use 3D volumes for depth inference. The quality of the reconstructed depth maps and the corresponding point clouds is directly influenced by the spatial resolution of the 3D volume. Consequently, these methods produce point clouds with sparse local regions because of the lack of the memory required to encode a high volume of information. Here, we apply the atrous spatial pyramid pooling (ASPP) module in MVS methods to obtain dense feature maps with multiscale, long-range, contextual information using high receptive fields. For a given 3D volume with the same spatial resolution as that in the MVS methods, the dense feature maps from the ASPP module encoded with superior information can produce dense point clouds without a high memory footprint. Furthermore, we propose a 3D loss for training the MVS networks, which improves the predicted depth values by 24.44%. The ASPP module provides state-of-the-art qualitative results by constructing relatively dense point clouds, which improves the DTU MVS dataset benchmarks by 2.25% compared with those achieved in the previous MVS methods.

• Proceedings of the Korea Water Resources Association Conference
• /
• 2022.05a
• /
• pp.143-143
• /
• 2022

Precipitation plays an essential role in water resources management and disaster prevention. Therefore, the understanding related to spatiotemporal characteristics of rainfall is necessary. Nowadays, highly accurate precipitation is mainly obtained from gauge observation systems. However, the density of gauge stations is a sparse and uneven distribution in mountainous areas. With the proliferation of technology, satellite-based precipitation sources are becoming increasingly common and can provide rainfall information in regions with complex topography. Nevertheless, satellite-based data is that it still remains uncertain. To overcome the above limitation, this study aims to take the strengthens of machine learning to generate a new reanalysis of precipitation data by fusion of multiple satellite precipitation products (SPPs) with gauge observation data. Several machine learning algorithms (i.e., Random Forest, Support Vector Regression, and Artificial Neural Network) have been adopted. To investigate the robustness of the new reanalysis product, observed data were collected to evaluate the accuracy of the products through Kling-Gupta efficiency (KGE), probability of detection (POD), false alarm rate (FAR), and critical success index (CSI). As a result, the new precipitation generated through the machine learning model showed higher accuracy than original satellite rainfall products, and its spatiotemporal variability was better reflected than others. Thus, reanalysis of satellite precipitation product based on machine learning can be useful source input data for hydrological simulations in ungauged river basins.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.18 no.1
• /
• pp.1-14
• /
• 2024

Time-series forecasting is extensively used in the actual world. Recent research has shown that Transformers with a self-attention mechanism at their core exhibit better performance when dealing with such problems. However, most of the existing Transformer models used for time series prediction use the traditional encoder-decoder architecture, which is complex and leads to low model processing efficiency, thus limiting the ability to mine deep time dependencies by increasing model depth. Secondly, the secondary computational complexity of the self-attention mechanism also increases computational overhead and reduces processing efficiency. To address these issues, the paper designs an efficient multi-layer attention-based time-series forecasting model. This model has the following characteristics: (i) It abandons the traditional encoder-decoder based Transformer architecture and constructs a time series prediction model based on multi-layer attention mechanism, improving the model's ability to mine deep time dependencies. (ii) A cross attention module based on cross attention mechanism was designed to enhance information exchange between historical and predictive sequences. (iii) Applying a recently proposed sparse attention mechanism to our model reduces computational overhead and improves processing efficiency. Experiments on multiple datasets have shown that our model can significantly increase the performance of current advanced Transformer methods in time series forecasting, including LogTrans, Reformer, and Informer.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.18 no.7
• /
• pp.2047-2066
• /
• 2024

In order to improve the coverage and achieve massive spectrum access, non-orthogonal multiple access (NOMA) technology is applied in millimeter wave massive multiple-input multiple-output (mMIMO) communication network. However, the power assumption of active sensors greatly limits its wide applications. Recently, Intelligent Reconfigurable Surface (IRS) technology has received wide attention due to its ability to reduce power consumption and achieve passive transmission. In this paper, spectral efficiency maximum problem in the millimeter wave mMIMO-NOMA system with IRS is considered. The sparse RF chain antenna structure is designed at the base station based on continuous phase modulation. Furthermore, a joint optimization problem for power allocation, power splitting, analog precoding and IRS reconfigurable matrices are constructed, which aim to achieve the maximum spectral efficiency of the system under the constraints of user's quality of service, minimum energy harvesting and total transmit power. A three-stage iterative algorithm is proposed to solve the above mentioned non-convex optimization problems. We obtain the local optimal solution by fixing some optimization parameters firstly, then introduce the relaxation variables to realize the global optimal solution. Simulation results show that the spectral efficiency of the proposed scheme is superior compared to the conventional system with phase shifter modulation. It is also demonstrated that IRS can effectively assist mmWave communication and improve the system spectral efficiency.

• Journal of the Institute of Electronics and Information Engineers
• /
• v.54 no.8
• /
• pp.81-88
• /
• 2017

This paper proposes a method of detecting rain regions from a single image. More specifically, a way of training the deep convolutional neural network based on the collected rain and non-rain patches is presented in a supervised manner. It is also shown that the proposed rain detection method based on deep convolutional neural network can provide better performance than the conventional rain detection method based on dictionary learning. Moreover, it is confirmed that the application of the proposed rain detection for rain removal can lead to some improvement in detail representation on the low-frequency regions of the rain-removed images. Additionally, this paper introduces the rain transfer method that inserts rain patterns into original images, thereby producing rain effects on the resulting images. The proposed rain transfer method could be used to augment rain patterns while constructing rain database.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.9 no.5
• /
• pp.1569-1586
• /
• 2015

This paper considers the optimal cell association and resource allocation for load balancing in a heterogeneous cellular network subject to user's quality-of-service (QoS) constraints. We adopt the proportional fairness (PF) utility maximization formulation which also accommodates the QoS constraints in terms of minimum rate requirements. With equal resource allocation this joint optimization problem is either infeasible or requires relaxation that yields a solution which is difficult to implement. Nevertheless, we show that this joint optimization problem can be effectively solved without any priori assumption on resource allocation and yields a cell association scheme which enforces single BS association for each user. We re-formulated the joint optimization problem as a network-wide resource allocation problem with cardinality constraints. A reweighted heuristic l₁-norm regularization method is used to obtain a sparse solution to the re-formulated problem. The cell association scheme is then derived from the sparsity pattern of the solution, which guarantees a single BS association for each user. Compared with the previously proposed method based on equal resource allocation, the proposed framework results in a feasible cell association scheme and yields a robust solution on resource allocation that satisfies the QoS constraints. Our simulations illustrate the impact of user's minimum rate requirements on cell association and demonstrate that the proposed approach achieves load balancing and enforces single BS association for users.

• Journal of Korea Water Resources Association
• /
• v.31 no.4
• /
• pp.439-448
• /
• 1998

There are many methods to calculate steady-state flowrate in a large water distribution system. Linear method which analyzes continuity equations and energy equations simultaneously is most widely used. Though it is theoretically simple, when it is applied to a practical water distribution system, it produces a very sparse coefficient matrix and most of its diagonal elements are to be zero. This sparsity characteristic of coefficient matrix makes it difficult to analyze pipe flow using the linear method. In this study, a graph theory is introduced to water distribution system analysis in order to prevent from producing ill-conditioned coefficient matrix and the technique is developed to produce positive-definite matrix. To test applicability of developed method, this method is applied to 22 pipes and 142 pipes system located nearby Taegu city. The results obtained from these applications show that the method can calculate flowrate effectively without failure in converage. Thus it is expected that the method can analyze steady state flowrate and pressure in pipe network systems efficiently. Keywords : pipe flow analysis, graph theory, linear method.

• The Journal of Korean Institute of Communications and Information Sciences
• /
• v.35 no.1B
• /
• pp.68-77
• /
• 2010

Wireless sensor networks are installed in various environments and collect sensing data through wireless links. The quality of a wireless link may be unstable due to environment causes and hardware performance in wireless sensor networks. Since the change of the link quality may cause data loss, sensor nodes need to adaptively estimate the change of the link quality. Also, the routing protocol should deal with this situation. In this paper, the adaptive link quality estimation and routing scheme in the large-scale wireless sensor networks are proposed. When the quality of a link is unstable, sensor nodes agilely estimate the quality of links, and the new route is selected. When quality of a link is stable, the link quality is occasionally estimated so that the energy consumption is reduced. Moreover, sensor nodes exchange less beacons in order to reduce an overhead in dense networks. In the case of sparse network, the sensor nodes exchange more beacons for finding a better route. We prove that the proposed scheme can improve the energy efficiency and reliability.

• Journal of Neurogastroenterology and Motility
• /
• v.24 no.4
• /
• pp.512-527
• /
• 2018

This review provides a comprehensive overview of brain imaging studies of the brain-gut interaction in functional gastrointestinal disorders (FGIDs). Functional neuroimaging studies during gut stimulation have shown enhanced brain responses in regions related to sensory processing of the homeostatic condition of the gut (homeostatic afferent) and responses to salience stimuli (salience network), as well as increased and decreased brain activity in the emotional response areas and reduced activation in areas associated with the top-down modulation of visceral afferent signals. Altered central regulation of the endocrine and autonomic nervous responses, the key mediators of the brain-gut axis, has been demonstrated. Studies using resting-state functional magnetic resonance imaging reported abnormal local and global connectivity in the areas related to pain processing and the default mode network (a physiological baseline of brain activity at rest associated with self-awareness and memory) in FGIDs. Structural imaging with brain morphometry and diffusion imaging demonstrated altered gray- and white-matter structures in areas that also showed changes in functional imaging studies, although this requires replication. Molecular imaging by magnetic resonance spectroscopy and positron emission tomography in FGIDs remains relatively sparse. Progress using analytical methods such as machine learning algorithms may shift neuroimaging studies from brain mapping to predicting clinical outcomes. Because several factors contribute to the pathophysiology of FGIDs and because its population is quite heterogeneous, a new model is needed in future studies to assess the importance of the factors and brain functions that are responsible for an optimal homeostatic state.

• Journal of Advanced Navigation Technology
• /
• v.27 no.2
• /
• pp.190-196
• /
• 2023

Recently, following the development of LIDAR technology which can detect distance from the object, the interest for LIDAR based 3D object detection network is getting higher. Previous networks generate inaccurate localization results due to spatial information loss during voxelization and downsampling. In this study, we propose an attention-based convergence method and a camera-LIDAR convergence system to acquire high-level features and high positional accuracy. First, by introducing the attention method into the Voxel-RCNN structure, which is a grid-based 3D object detection network, the multi-scale sparse 3D convolution feature is effectively fused to improve the performance of 3D object detection. Additionally, we propose the late-fusion mechanism for fusing outcomes in 3D object detection network and 2D object detection network to delete false positive. Comparative experiments with existing algorithms are performed using the KITTI data set, which is widely used in the field of autonomous driving. The proposed method showed performance improvement in both 2D object detection on BEV and 3D object detection. In particular, the precision was improved by about 0.54% for the car moderate class compared to Voxel-RCNN.