• Journal of Korea Multimedia Society
• /
• v.24 no.3
• /
• pp.416-422
• /
• 2021

Due to the large differences in human activity within classes, the large similarity between classes, and the problems of visual angle and occlusion, it is difficult to extract features manually, and the detection rate of human behavior is low. In order to better solve these problems, an improved Faster R-CNN-based detection algorithm is proposed in this paper. It achieves multi-object recognition and localization through a second-order detection network, and replaces the original feature extraction module with Dense-Net, which can fuse multi-level feature information, increase network depth and avoid disappearance of network gradients. Meanwhile, the proposal merging strategy is improved with Soft-NMS, where an attenuation function is designed to replace the conventional NMS algorithm, thereby avoiding missed detection of adjacent or overlapping objects, and enhancing the network detection accuracy under multiple objects. During the experiment, the improved Faster R-CNN method in this article has 84.7% target detection result, which is improved compared to other methods, which proves that the target recognition method has significant advantages and potential.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.16 no.9
• /
• pp.2882-2903
• /
• 2022

Image super-resolution (SR) processing is of great value in the fields of digital image processing, intelligent security, film and television production and so on. This paper proposed a densely connected deep learning network based on cascade architecture, which can be used to solve the problem of super-resolution in the field of image quality enhancement. We proposed a more efficient residual scaling dense block (RSDB) and the multi-channel cascade architecture to realize more efficient feature reuse. Also we proposed a hybrid loss function based on L₁ error and L_∞ error to achieve better L_∞ error performance. The experimental results show that the overall performance of the network is effectively improved on cascade architecture and residual scaling. Compared with the residual dense net (RDN), the PSNR / SSIM of the new method is improved by 2.24% / 1.44% respectively, and the L_∞ performance is improved by 3.64%. It shows that the cascade connection and residual scaling method can effectively realize feature reuse, improving the residual convergence speed and learning efficiency of our network. The L_∞ performance is improved by 11.09% with only a minimal loses of 1.14% / 0.60% on PSNR / SSIM performance after adopting the new loss function. That is to say, the L_∞ performance can be improved greatly on the new loss function with a minor loss of PSNR / SSIM performance, which is of great value in L_∞ error sensitive tasks.

• The Transactions of The Korean Institute of Electrical Engineers
• /
• v.63 no.4
• /
• pp.451-454
• /
• 2014

Short-term load forecasting(STLF) for electric power demand is essential for stable power system operation and efficient power market operation. We improved STLF method by using hourly temperature as an input data. In order to using hourly temperature to STLF algorithm, we calculated temperature-electric power demand sensitivity through past actual data and combined this sensitivity to exponential smoothing method which is one of the STLF method. The proposed method is verified by case study for a week. The result of case study shows that the average percentage errors of the proposed load forecasting method are improved comparing with errors of the previous methods.

• Journal of Institute of Control, Robotics and Systems
• /
• v.22 no.9
• /
• pp.753-758
• /
• 2016

In this paper, we propose a novel method for estimating a vehicle's current position, even on roads that have similar patterns. In the proposed method, we classified the semantic information of the nodes in detail and added the semantic information of the link to solve the problem due to similar and repeated patterns. We also improved the mapping method by comparing the result of the duplicated matching with that of the only matching obtained just before corresponding duplicated matching. From the simulation results, we verify that the performance of the proposed method is better than that of the existing method.

• Journal of Power Electronics
• /
• v.18 no.5
• /
• pp.1608-1617
• /
• 2018

The rapid and correct isolation of faulty submodules (SMs) is of great importance for improving the reliability of modular multilevel converters (MMCs). Therefore, a fast diagnosis method containing fault detection and fault location determination was presented in this paper. An improved incremental predictive model of arm current was proposed to detect failures, and the multi-step prediction method was used to eliminate the negative impact of disturbances. Moreover, a control method was proposed to strengthen the fault characteristics to rapidly locate faulty arms and faulty SMs by detecting the variation rate of the SM capacitor voltage. The proposed method can rapidly and easily locate faulty SMs under different load conditions without the need for additional sensors. The experimental results have validated the effectiveness of the proposed method by using a single-phase MMC with four SMs per arm.

• Proceedings of the KIEE Conference
• /
• 2001.07b
• /
• pp.1035-1037
• /
• 2001

This paper presents an improved hys- teresis current control method for three-phase PWM power inverters using 3-level comparator. Hysteresis current controller using 3-level comparator has an advantage of constant switching frequency compared with conventional hysteresis current controller. However, this method has disadvantage that the longer sampling period, the larger current error because the switching is performed without considering current error magnitude of each phase. The proposed method improves the control performance by selecting the optimum switching pattern in which the magnitudes of current errors are considered introducing space vector concept. Simulation results using Matlab/Simulink show that the proposed control method reduces current error keeping the merit of previous hysteresis current control method.

• Proceedings of the Korea Concrete Institute Conference
• /
• 2009.05a
• /
• pp.569-570
• /
• 2009

An improvement of the test method is proposed to more accurately measure early-age autogenous shrinkage in concrete particularly within first 24 hours after casting. Experiments were conducted considering existing and improved method. In improved method, hydration temperature was artificially controlled to prevent thermal deformations. Test results indicate that the autogenous shrinkage calculated by existing approach is underestimated which might be due to the wrong assumption of considering the thermal dilation coefficient to be constant (equal to 10 ${\times}$ $10^{-6}/^{\circ}C$) at early-age. We recommend that the proposed method should be adopted to better assess precise value of autogenous shrinkage or an appropriate method of determining the time-evolution of thermal dilation coefficient be considered.

• Journal of the Korean Institute of Telematics and Electronics B
• /
• v.32B no.12
• /
• pp.1614-1623
• /
• 1995

In this paper we propose a region-adaptive CELP image coder for still images at low bit rates below 0.5 bpp. The proposed method partitions the image into stochastically similar regions by the minimum spanning tree method and finds prediction coefficients for each region using a 2- dimensional linear prediction model. Coding is carried out on 8$\times$8 blocks and when there are several regions included in a block, an image is synthesized using the prediction coefficients of each region. Computer simulation results show that the proposed method allows improved synthesized image over conventional block-adaptive CELP methods, especially at edges. In addition, performance comparison with the JPEG DCT method shows that while the JPEG method shows block distortion and staircase effects (ragged edges) at bit rates below 0.5 bpp, the proposed CELP method shows improved synthesized images with such effects reduced.

• ETRI Journal
• /
• v.29 no.1
• /
• pp.18-26
• /
• 2007

This paper presents an improved redundant picture coding method that efficiently enhances the error resiliency of H.264. The proposed method applies polyphase downsampling to residual blocks obtained from inter prediction and selectively encodes the rearranged residual blocks in the redundant picture coding process. Moreover, a spatial-temporal sample construction method is developed for the redundant coded picture, which further improves the reconstructed picture quality in error prone environments. Simulations based on JM11.0 were run to verify the proposed method on different test sequences in various error prone environments with average packet loss rates of 3%, 5%, 10%, and 20%. Results of the simulations show that the presented method significantly improves the robustness of H.264 to packet loss by 1.6 dB PSNR on average over the conventional redundant picture coding method.

• Journal of the Korean Institute of Telematics and Electronics S
• /
• v.34S no.3
• /
• pp.55-62
• /
• 1997

For the effective removal of artifacts and the extraction of an improved evoked potential response, we propose the averaging method usin gthe shrinkag eof wavelet coefficients. The wavelet analysis decomposes the measured evoked potentials into scale coefficients with low frequency components and wavelet coefficients with high ones as a resolution level, respectively. and in the course of synthesis evoked potentials, the presented method shrinks the wavelet coefficients, and then reproduces the evoked potentials, and lastly averages it. We measured visual evoked potentials to simulate the averaging method using the shrinkage of wavelet coefficients, and compared it with aveaged signal. As a result of simulations, the proposed method gets improved VEP about 0.2-1.6dB in comparison with the averaging method with daubechies wavelet in the resolution level four.