• The Transactions of the Korea Information Processing Society
• /
• v.13 no.9
• /
• pp.444-452
• /
• 2024

This study proposes a deep learning architecture optimized for fire detection derived through Layer Importance Evaluation. In order to solve the problem of unnecessary complexity and operation of the existing Convolutional Neural Network (CNN)-based fire detection system, the operation of the inner layer of the model based on the weight and activation values was analyzed through the Layer Importance Evaluation technique, the layer with a high contribution to fire detection was identified, and the model was reconstructed only with the identified layer, and the performance indicators were compared and analyzed with the existing model. After learning the fire data using four transfer learning models: Xception, VGG19, ResNet, and EfficientNetB5, the Layer Importance Evaluation technique was applied to analyze the weight and activation value of each layer, and then a new model was constructed by selecting the top rank layers with the highest contribution. As a result of the study, it was confirmed that the implemented architecture maintains the same performance with parameters that are about 80% lighter than the existing model, and can contribute to increasing the efficiency of fire monitoring equipment by outputting the same performance in accuracy, loss, and confusion matrix indicators compared to conventional complex transfer learning models while having a learning speed of about 3 to 5 times faster.

• Journal of the Korean Society of Surveying, Geodesy, Photogrammetry and Cartography
• /
• v.31 no.6_2
• /
• pp.559-565
• /
• 2013

Spatial resolution of land covers from remotely sensed images can be increased using super-resolution mapping techniques for soft-classified land cover proportions. A further development of super-resolution mapping technique is downscaling the original remotely sensed image using super-resolution mapping techniques with a forward model. In this paper, the model for increasing spatial resolution of remote sensing multispectral image is tested with real SPOT 5 imagery at 10m spatial resolution for an area in Bac Giang Province, Vietnam in order to evaluate the feasibility of application of this model to the real imagery. The soft-classified land cover proportions obtained using a fuzzy c-means classification are then used as input data for a Hopfield neural network (HNN) to predict the multispectral images at sub-pixel spatial resolution. The 10m SPOT multispectral image was improved to 5m, 3,3m and 2.5m and compared with SPOT Panchromatic image at 2.5m resolution for assessment.Visually, the resulted image is compared with a SPOT 5 panchromatic image acquired at the same time with the multispectral data. The predicted image is apparently sharper than the original coarse spatial resolution image.

• Journal of Institute of Control, Robotics and Systems
• /
• v.3 no.2
• /
• pp.204-213
• /
• 1997

In this paper, an application of fuzzy-neuron reasoning to the control of an activated sludge plant is presented. The activated sludge process is widely used in modern wastewater treatment plants. The operation control of the activated sludge process, however, is difficult due to the following reasons : 1)The complexity of the wastewater components, 2)the change of the wastewater influent, and 3)the adjustment errors in the control process. Because of these reasons, it is difficult to obtain mathematical model that really reflect the relationship between the variables and parameters in the process of wastewater treatment correctively and effectively. In this paper, the activated sludge process(A.S.P.) is modeled by a new fuzzy-neuron network representing nonlinear characteristics. These fuzzy-neurons have fuzzy rules with complementary membership function. Based on the constructed model, graphic simulator on X-window system as a graphic integrated environment is implemented. The efficacy of the proposed control scheme was evaluated and demonstrated by means of the field test.

• Journal of Communications and Networks
• /
• v.16 no.1
• /
• pp.36-44
• /
• 2014

Exploiting the energy-delay tradeoff for energy saving is critical for developing green wireless communication systems. In this paper, we investigate the delay-constrained energy-efficient packet transmission. We aim to minimize the energy consumption of multiple randomly arrived packets in an additive white Gaussian noise channel subject to individual packet delay constraints, by taking into account the practical on-off circuit power consumption at the transmitter. First, we consider the offline case, by assuming that the full packet arrival information is known a priori at the transmitter, and formulate the energy minimization problem as a non-convex optimization problem. By exploiting the specific problem structure, we propose an efficient scheduling algorithm to obtain the globally optimal solution. It is shown that the optimal solution consists of two types of scheduling intervals, namely "selected-off" and "always-on" intervals, which correspond to bits-per-joule energy efficiency maximization and "lazy scheduling" rate allocation, respectively. Next, we consider the practical online case where only causal packet arrival information is available. Inspired by the optimal offline solution, we propose a new online scheme. It is shown by simulations that the proposed online scheme has a comparable performance with the optimal offline one and outperforms the design without considering on-off circuit power as well as the other heuristically designed online schemes.

• Journal of Communications and Networks
• /
• v.16 no.1
• /
• pp.45-55
• /
• 2014

The paper investigates the problem of designing the multiple-antenna relay in a two-way relay network by taking into account the imperfect channel state information (CSI). The objective is to design the multiple-antenna relay based upon the CSI estimates, where the estimation errors are included to attain the robust design under the worst-case philosophy. In particular, the worst-case transmit power at the multiple-antenna relay is minimized while guaranteeing the worst-case quality of service requirements that the received signal-to-noise ratio (SNR) at both sources are above a prescribed threshold value. Since the worst-case received SNR expression is too complex for subsequent derivation and processing, its lower bound is explored instead by minimizing the numerator and maximizing the denominator of the worst-case SNR. The aforementioned problem is mathematically formulated and shown to be nonconvex. This motivates the pursuit of semidefinite relaxation coupled with a randomization technique to obtain computationally efficient high-quality approximate solutions. This paper has shown that the original optimization problem can be reformulated and then relaxed to a convex problem that can be solved by utilizing suitable randomization loop. Numerical results compare the proposed multiple-antenna relay with the existing nonrobust method, and therefore validate its robustness against the channel uncertainty. Finally, the feasibility of the proposed design and the associated influencing factors are discussed by means of extensive Monte Carlo simulations.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.14 no.2
• /
• pp.855-870
• /
• 2020

Windows are key components of building facades. Detecting windows, crucial to 3D semantic reconstruction and scene parsing, is a challenging task in computer vision. Early methods try to solve window detection by using hand-crafted features and traditional classifiers. However, these methods are unable to handle the diversity of window instances in real scenes and suffer from heavy computational costs. Recently, convolutional neural networks based object detection algorithms attract much attention due to their good performances. Unfortunately, directly training them for challenging window detection cannot achieve satisfying results. In this paper, we propose an approach for window detection. It involves an improved Faster R-CNN architecture for window detection, featuring in a window region proposal network, an RoI feature fusion and a context enhancement module. Besides, a post optimization process is designed by the regular distribution of windows to refine detection results obtained by the improved deep architecture. Furthermore, we present a newly collected dataset which is the largest one for window detection in real street scenes to date. Experimental results on both existing datasets and the new dataset show that the proposed method has outstanding performance.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.12 no.4
• /
• pp.1674-1692
• /
• 2018

An RFID network is a network in which a reader inquire about the identities of tags and tags respond with their identities to a reader. The diversity of RFID networks has brought about many applications including an inexpensive system where a single reader supports a small number of tags. Such a system needs a tag cognizance scheme that is able to arbitrate among contending tags as well as is simple enough. In this paper, confining our attention to a clan of simple schemes, we propose a randomized scheme with aiming at enhancing the tag cognizance rate than a conventional scheme. Then, we derive an exact expression for the cognizance rate attained by the randomized scheme. Unfortunately, the exact expression is not so tractable as to optimize the randomized scheme. As an alternative way, we develop an upper bound on the tag cognizance rate. In a closed form, we then obtain a nearly optimal value for a key design parameter, which maximizes the upper bound. Numerical examples confirm that the randomized scheme is able to dominate the conventional scheme in cognizance rate by employing a nearly optimal value. Furthermore, they reveal that the randomized scheme is robust to the fallacy that the reader believes or guesses a wrong number of neighboring tags.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.9 no.3
• /
• pp.1121-1139
• /
• 2015

In texture-plus-depth format, depth map compression is an important task. Different from normal texture images, depth maps have less texture information, while contain many homogeneous regions separated by sharp edges. This feature will be employed to form an efficient depth map coding scheme in this paper. Firstly, the histogram of the depth map will be analyzed to find an appropriate threshold that segments the depth map into the foreground and background regions, allowing the edge between these two kinds of regions to be obtained. Secondly, the two regions will be encoded through rate distortion optimization with a shape adaptive wavelet transform, while the edges are lossless encoded with JBIG2. Finally, a depth-updating algorithm based on the threshold and the depth range is applied to enhance the quality of the decoded depth maps. Experimental results demonstrate the effective performance on both the depth map quality and the synthesized view quality.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.3 no.5
• /
• pp.423-443
• /
• 2009

We present the design and implementation of a unique sensing and actuation application -- the Illuminator: a sensor network-based intelligent light control system for entertainment and media production. Unlike most sensor network applications, which focus on sensing alone, a distinctive aspect of the Illuminator is that it closes the loop from light sensing to lighting control. We describe the Illuminator's design requirements, system architecture, algorithms, implementation and experimental results. The system uses the Illumimote, a multi-modal and high fidelity light sensor module well-suited for wireless sensor networks, to satisfy the high-performance light sensing requirements of entertainment and media production applications. The Illuminator system is a toolset to characterize the illumination profile of a deployed set of fixed position lights, generate desired lighting effects for moving targets (actors, scenic elements, etc.) based on user constraints expressed in a formal language, and to assist in the set up of lights to achieve the same illumination profile in multiple venues. After characterizing deployed lights, the Illuminator computes optimal light settings at run-time to achieve a user-specified actuation profile, using an optimization framework based on a genetic algorithm. Uniquely, it can use deployed sensors to incorporate changing ambient lighting conditions and moving targets into actuation. Experimental results demonstrate that the Illuminator handles various high-level user requirements and generates an optimal light actuation profile. These results suggest that the Illuminator system supports entertainment and media production applications.

• Journal of Electrical Engineering and Technology
• /
• v.4 no.3
• /
• pp.315-322
• /
• 2009

Penetration and installation of a new dynamic technology known as Flexible AC Transmission Systems (FACTS) in a practical and dynamic network requires and force expert engineer to develop robust and flexible strategy for planning and control. Unified Power Flow Controller (UPFC) is one of the recent and effective FACTS devices designed for multi control operation to enhance the power system security. This paper presents a dynamic strategy based on Particle Swarm Optimization (PSO) for optimal parameters setting of UPFC to enhance the system loadability. Firstly, we perform a multi power flow analysis with load incrementation to construct a global database to determine the initial efficient bounds associated to active power and reactive power target vector. Secondly a PSO technique applied to search the new parameters setting of the UPFC within the initial new active power and reactive power target bounds. The proposed approach is implemented with Matlab program and verified with IEEE 30-Bus test network. The results show that the proposed approach can converge to the near optimum solution with accuracy, and confirm that flexible multi-control of this device coordinated with efficient location enhance the system security of power system by eliminating the overloaded lines and the bus voltage violation.