• Journal of Information Processing Systems
• /
• v.18 no.1
• /
• pp.12-25
• /
• 2022

In recent years, deep convolutional neural networks have made significant progress in the research of single image super-resolution. However, it is difficult to be applied in practical computing terminals or embedded devices due to a large number of parameters and computational effort. To balance these problems, we propose CSRNet, a lightweight neural network based on channel split residual learning structure, to reconstruct highresolution images from low-resolution images. Lightweight refers to designing a neural network with fewer parameters and a simplified structure for lower memory consumption and faster inference speed. At the same time, it is ensured that the performance of recovering high-resolution images is not degraded. In CSRNet, we reduce the parameters and computation by channel split residual learning. Simultaneously, we propose a double-upsampling network structure to improve the performance of the lightweight super-resolution network and make it easy to train. Finally, we propose a new evaluation metric for the lightweight approaches named 100_FPS. Experiments show that our proposed CSRNet not only speeds up the inference of the neural network and reduces memory consumption, but also performs well on single image super-resolution.

• Journal of Institute of Control, Robotics and Systems
• /
• v.5 no.1
• /
• pp.62-68
• /
• 1999

The attitude of a vehicle can be precisely determined using GPS carrier phase measurements from more than two antennas attached to a vehicle and an efficient integer ambiguity resolution technique. Many methods utilizing the known baseline length as a constraint of independent elements of integer ambiguities are proposed to resolve integer ambiguity at real time. Three-dimensional search space is reduced to two-dimensional search space with this constraint. Thus the true integer ambiguity can be easily determined with less computational burden and fewer number of measurements. But there are still strong requirements for the real time integer ambiguity resolution, which uses single epoch measurement of long baseline. In this paper, a new constraint from the geometry of multiple baselines is derived. With this new constraint, two-dimensional search space is further reduced to one-dimensional search space. It makes possible to determine integer ambiguity with single epoch measurement. The proposed method is applied to real data to show its effectiveness.

• Journal of Biomedical Engineering Research
• /
• v.22 no.3
• /
• pp.223-229
• /
• 2001

In this paper, we have compared the direct fan-beam reconstruction method with the rebinning method in terms of computation time and spatial resolution using computer simulation. As a result, the direct fan-beam method is superior to the rebinning method in the spatial resolution though the former needs longer computation time. However, if we adopt the quarter-detector-offset technique to improve the spatial resolution, the rebinning method outperforms the direct fan-beam method. The computation times have been evaluated using the fast algorithms optimized to reduce the number of interpolation calculations at the back-projection, and the spatial resolutions have been compared using the computer generated phantoms.

• Journal of electromagnetic engineering and science
• /
• v.15 no.1
• /
• pp.37-43
• /
• 2015

We implemented a real-time radar system for an unmanned ground vehicle designed to run on unpaved or bumpy roads. The system must be able to detect slow targets in a cluttered environment and cover wide angular sections with high resolution at the same time. The system consists of array antennas, preprocessors for digital beam forming, and digital signal processors for the detection process which uses sawtooth waveforms and high-resolution estimation, and is called forward/backward spatial smoothing beamspace multiple signal classification (FBSS BS-MUSIC). We show that the sawtooth waveforms enhance the angular estimation capability of FBSS BS-MUSIC in addition to their well-known advantages of removing the ambiguity of targets and detecting slow targets with improved velocity resolution.

• Korean Journal of Optics and Photonics
• /
• v.19 no.5
• /
• pp.370-375
• /
• 2008

High-speed high-resolution terahertz time-domain spectroscopy (THz-TDS) is demonstrated using the asynchronous-opticalsampling (AOS) method. A time-domain signal with a 10-ns time window is rapidly acquired by using two femtosecond lasers with slightly different repetition frequencies to generate and detect a terahertz pulse wave, without a mechanical delay stage. The spectrum obtained by the fast Fourier transformation (FFT) of the time-domain waveform has a frequency resolution of 100 MHz. The time resolution of our spectrometer is measured using the cross-correlation method to be 278 fs. A transmission spectrum of water vapor is measured and the absorption lines are analyzed in the frequency range from 0.1 to 1.2 THz.

• The Journal of Korean Institute of Electromagnetic Engineering and Science
• /
• v.7 no.4
• /
• pp.328-335
• /
• 1996

The numerical dispersive characteristics and the numerical stability confition of the Multi-Resolution Time-Domain(MRTD) method are calculated. A dispersion analysis of the MRTD schemes including a comparison to Yee's Finite-Difference Time-Domain(FDTD) method is given. The superiority of the MRTD method to the spatial discretization is shown. The required computational memory can be reduced by using the MRTD method. We expect that the MRTD method will be very useful method for numerical modelling of electromagnetics.

• Proceedings of the Korean Society of Precision Engineering Conference
• /
• 2006.10a
• /
• pp.81-82
• /
• 2006

• Journal of the Institute of Electronics and Information Engineers
• /
• v.54 no.8
• /
• pp.12-19
• /
• 2017

In this paper a super-resolution time delay estimation algorithm that estimates the time delays of spread spectrum signals using sparse signal reconstruction approach is introduced. So far, the correlation method has been mostly used to estimate the time delays of spread spectrum signals. However it fails to accurately estimate the time delays in the case where the signals are spaced within approximately 1 PN chip duration and a further processing should be applied to the correlation outputs in order to enhance the resolution capability. Recently sparse signal approaches attract much interest in the area of directions-of-arrival estimation, of which SPICE is the most representative. Thus we introduce a super-resolution time delay estimation algorithm based on the SPICE approach and compare its performance with that of MUSIC algorithm by applying them to the ISO/IEC 24730-2.1 RTLS system.

• ETRI Journal
• /
• v.37 no.2
• /
• pp.251-261
• /
• 2015

We present a novel method for real-time automatic license plate detection in high-resolution videos. Although there have been extensive studies of license plate detection since the 1970s, the suggested approaches resulting from such studies have difficulties in processing high-resolution imagery in real-time. Herein, we propose a novel cascade structure, the fastest classifier available, by rejecting false positives most efficiently. Furthermore, we train the classifier using the core patterns of various types of license plates, improving both the computation load and the accuracy of license plate detection. To show its superiority, our approach is compared with other state-of-the-art approaches. In addition, we collected 20,000 images including license plates from real traffic scenes for comprehensive experiments. The results show that our proposed approach significantly reduces the computational load in comparison to the other state-of-the-art approaches, with comparable performance accuracy.

• Proceedings of the Korean Society of Agricultural Engineers Conference
• /
• 2002.10a
• /
• pp.453-456
• /
• 2002

In this study, the effect of DEM resolution (15m, 30m, 50m, 70m, 100m, 200m, 300m) on the hydrological simulation was examined using BASINS (Better Assessment Science Integrating point and Nonpoint Source) for Heukcheon watershed (303.3km2) data from 1998 to 1999. Generally, as the cell size of DEM increased, topographical changes were observed as the original range of elevation decreased. The processing time of watershed delineation and river network needed more time and effort on smaller cell size of DEM. The larger DEM demonstrated had some errors in the junction of river network which might effects on the simulation of water quantity and quality. The area weighted average watershed slope became lower but the length weighted average channel slope became higher as the DEM size increased. DEM resolution affected substantially on the topographical parameter but less on the hydrological simulation. Considering processing time and accuracy on hydrological simulation DEM mesh size of 100m is recommended for this watershed.