• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.11 no.10
• /
• pp.4968-4986
• /
• 2017

In this paper, we primarily address the difficulty of automatic generation of a plausible depth map from a single image in an unstructured environment. The aim is to extrapolate a depth map with a more correct, rich, and distinct depth order, which is both quantitatively accurate as well as visually pleasing. Our technique, which is fundamentally based on a preexisting DepthTransfer algorithm, transfers depth information at the level of superpixels. This occurs within a framework that replaces a pixel basis with one of instance-based learning. A vital superpixels feature enhancing matching precision is posterior incorporation of predictive semantic labels into the depth extraction procedure. Finally, a modified Cross Bilateral Filter is leveraged to augment the final depth field. For training and evaluation, experiments were conducted using the Make3D Range Image Dataset and vividly demonstrate that this depth estimation method outperforms state-of-the-art methods for the correlation coefficient metric, mean log10 error and root mean squared error, and achieves comparable performance for the average relative error metric in both efficacy and computational efficiency. This approach can be utilized to automatically convert 2D images into stereo for 3D visualization, producing anaglyph images that are visually superior in realism and simultaneously more immersive.

• Proceedings of the Korean Society of Medical Physics Conference
• /
• 2002.09a
• /
• pp.373-375
• /
• 2002

4D CT is a dynamic volume imaging system of moving organs with an image quality comparable to conventional CT, and is realized with continuous and high-speed cone-beam CT. In order to realize 4D CT, we have developed a novel 2D detector on the basis of the present CT technology, and mounted it on the gantry frame of the state-of-the-art CT-scanner. In the present report we describe the design of the first model of 4D CT-scanner as well as the early results of performance test. The x-ray detector for the 4D CT-scanner is a discrete pixel detector in which pixel data are measured by an independent detector element. The numbers of elements are 912 (channels) ${\times}$ 256 (segments) and the element size is approximately 1mm ${\times}$ 1mm. Data sampling rate is 900views(frames)/sec, and dynamic range of A/D converter is 16bits. The rotation speed of the gantry is l.0sec/rotation. Data transfer system between rotating and stationary parts in the gantry consists of laser diode and photodiode pairs, and achieves net transfer speed of 5Gbps. Volume data of 512${\times}$512${\times}$256 voxels are reconstructed with FDK algorithm by parallel use of 128 microprocessors. Normal volunteers and several phantoms were scanned with the scanner to demonstrate high image quality.

• Journal of the Institute of Electronics Engineers of Korea SP
• /
• v.46 no.5
• /
• pp.151-161
• /
• 2009

Audio Video Standard (AVS) is the audio and video compression standard that was developed for domestic video applications in China. AVS employs low complexity tools to minimize degradation of RD performance of the state-the-art video codec, H.264/AVC. The AVS video codec consists of $8{\times}8$ block prediction and the same size transform to improve compression efficiency for VGA and higher resolution sequences. Currently, the AVS has been adopted more and more for IPTV services and mobile applications in China. So, many consumer electronics companies and multimedia-related laboratories have been developing applications and chips for the AVS. In this paper, we implemented the AVS video decoder and optimize it on TI's Davinci EVM DSP board. For improving the decoding speed and clocks, we removed unnecessary memory operations and we also used high-speed VLD algorithm, linear assembly, intrinsic functions and so forth. Test results show that decoding speed of the optimized decoder is $5{\sim}7$ times faster than that of the reference software (RM 5.2J).

• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.12 no.6
• /
• pp.1095-1100
• /
• 2008

In a process of insect foot recognition, footprint segments as basic areas for recognition need to be extracted from scanned insect footprints and appropriate features should be found from the footprint segments in order to discriminate kinds of insects, because the characteristics of the features are important to classify insects. In this paper, we propose methods for automatic footprint segmentation and feature extraction. We use a Trace transform method in order to find out appropriate features from the extracted segments by the above methods. The Trace transform method builds a new type of data structure from the segmented images by functions using parallel trace lines and the new type of data structure has characteristics invariant to translation, rotation and reflection of images. This data structure is converted to Triple features by Diametric and Circus functions, and the Triple features are used for discriminating patterns of insect footprints. In this paper, we show that the Triple features found by the proposed methods are enough distinguishable and appropriate for classifying kinds of insects.

• Korean Journal of Remote Sensing
• /
• v.25 no.6
• /
• pp.547-557
• /
• 2009

LIDAR (LIght Detection And Ranging) is an active remote sensing technology which provides 3D coordinates of the Earth's surface by performing range measurements from the sensor. Early small footprint LIDAR systems recorded multiple discrete returns from the back-scattered energy. Recent advances in LIDAR hardware now make it possible to record full digital waveforms of the returned energy. LIDAR waveform decomposition involves separating the return waveform into a mixture of components which are then used to characterize the original data. The most common statistical mixture model used for this process is the Gaussian mixture. Waveform decomposition plays an important role in LIDAR waveform processing, since the resulting components are expected to represent reflection surfaces within waveform footprints. Hence the decomposition results ultimately affect the interpretation of LIDAR waveform data. Computational requirements in the waveform decomposition process result from two factors; (1) estimation of the number of components in a mixture and the resulting parameter estimates, which are inter-related and cannot be solved separately, and (2) parameter optimization does not have a closed form solution, and thus needs to be solved iteratively. The current state-of-the-art airborne LIDAR system acquires more than 50,000 waveforms per second, so decomposing the enormous number of waveforms is challenging using traditional single processor architecture. To tackle this issue, four parallel LIDAR waveform decomposition algorithms with different work load balancing schemes - (1) no weighting, (2) a decomposition results-based linear weighting, (3) a decomposition results-based squared weighting, and (4) a decomposition time-based linear weighting - were developed and tested with varying number of processors (8-256). The results were compared in terms of efficiency. Overall, the decomposition time-based linear weighting work load balancing approach yielded the best performance among four approaches.

• Korean Journal of Heritage :History & Science
• /
• v.50 no.2
• /
• pp.80-93
• /
• 2017

This study proposes a new scientific measurement method for damaged epigraph. In this new method, the Chinese characters are converted and coordinates are created for this measurement. This method is then used to decipher partially damaged characters from the parts of the coordinated characters that are damaged and intact. The Chinese characters are divided into 9 square parts by the position of their Chinese Radicals. The unknown characters are then compared and deciphered dependent upon the character shape in 9 square parts that have been created. This method is more scientific, accurate, and makes it easier to find related characters than deciphering through contexts, which is current method. When creating a new software based on this algorithm, it will be especially useful in deciphering an old manuscript or a epigraph that made ancient Chinese characters which are not currently in use. This study will also be helpful in deciphering semi-cursive styled or cursive styled epigraph, as well as semi-cursive styled or cursive styled damaged characters during follow-up research.

• International Journal of Advanced Culture Technology
• /
• v.9 no.2
• /
• pp.100-105
• /
• 2021

Contactless service is rapidly emerging as a new growth strategy due to consumers who are reluctant to the face-to-face situation in the global pandemic of coronavirus disease 2019 (COVID-19), and various technologies are being developed to support the fast-growing contactless service market. In particular, the restaurant industry is one of the most desperate industrial fields requiring technologies for contactless service, and the representative technical case should be a kiosk, which has the advantage of reducing labor costs for the restaurant owners and provides psychological relaxation and satisfaction to the customer. In this paper, we propose a solution to the restaurant's store operation through the unmanned kiosk using a state-of-the-art artificial intelligence (AI) technology of image recognition. Especially, for the products that do not have barcodes in bakeries, fresh foods (fruits, vegetables, etc.), and autonomous restaurants on highways, which cause increased labor costs and many hassles, our proposed system should be very useful. The proposed system recognizes products without barcodes on the ground of image-based AI algorithm technology and makes automatic payments. To test the proposed system feasibility, we established an AI vision system using a commercial camera and conducted an image recognition test by training object detection AI models using donut images. The proposed system has a self-learning system with mismatched information in operation. The self-learning AI technology allows us to upgrade the recognition performance continuously. We proposed a fully automated payment system with AI vision technology and showed system feasibility by the performance test. The system realizes contactless service for self-checkout in the restaurant business area and improves the cost-saving in managing human resources.

• Journal of the Institute of Electronics Engineers of Korea SP
• /
• v.49 no.2
• /
• pp.35-44
• /
• 2012

Most recent infrared (IR) sensors have a focal-plane array (FPA) structure. Spatial non-uniformity of a FPA structure, however, introduces unwanted fixed pattern noise (FPN) to images. This non-uniformity correction (NUC) of a FPA can be categorized into target-based and scene-based approaches. In a target-based approach, FPN can be separated by using a uniform target such as a black body. Since the detector response randomly drifts along the time axis, however, several scene-based algorithms on the basis of a video sequence have been proposed. Among those algorithms, the state-of-the-art one based on Kalman filter uses one-directional warping for motion compensation and only compensates for offset non-uniformity of IR camera detectors. The system model using one-directional warping cannot correct the boundary region where a new scene is being introduced in the next video frame. Furthermore, offset-only correction approaches may not completely remove the FPN in images if it is considerably affected by gain non-uniformity. Therefore, for FPN reduction in IR videos, we propose a joint correction algorithm of gain and offset based on bi-directional warping. Experiment results using simulated and real IR videos show that the proposed scheme can provide better performance compared with the state-of-the art in FPN reduction.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.12 no.4
• /
• pp.1504-1526
• /
• 2018

For uplink multi-user massive MIMO systems, conventional minimum mean square error (MMSE) linear detection method achieves near-optimal performance when the number of antennas at base station is much larger than that of the single-antenna users. However, MMSE detection involves complicated matrix inversion, thus making it cumbersome to be implemented cost-effectively and rapidly. In this paper, we first summarize in detail the state-of-the-art simplified MMSE detection algorithms that circumvent the complicated matrix inversion and hence reduce the computation complexity from ${\mathcal{O}}(K^3)$ to ${\mathcal{O}}(K^2)$ or ${\mathcal{O}}(NK)$ with some certain performance sacrifice. Meanwhile, we divide the simplified algorithms into two categories, namely the matrix inversion approximation and the classical iterative linear equation solving methods, and make comparisons between them in terms of detection performance and computation complexity. In order to further optimize the detection performance of the existing detection algorithms, we propose more proper solutions to set the initial values and relaxation parameters, and present a new way of reconstructing the exact effective noise variance to accelerate the convergence speed. Analysis and simulation results verify that with the help of proper initial values and parameters, the simplified matrix inversion based detection algorithms can achieve detection performance quite close to that of the ideal matrix inversion based MMSE algorithm with only a small number of series expansions or iterations.

• Journal of Korea Spatial Information System Society
• /
• v.11 no.2
• /
• pp.177-181
• /
• 2009

Developing u-Public facilities for application u-City is to combine both the state-of-the art of the construction and ubiquitous computing and must be flexibly comprised of the facilities for the basic service of the building such as air conditioning, heating, lighting and electric equipments to materialize a new format of spatial planning and the public facilities inside or outside. Accordingly, in this paper we suggested the time pattern system for predicting the most basic power system loads for the basic service. To application the tim e pattern we applied SOM algorithm and k-means method and then clustered the data each weekday and each time respectively. The performance evaluation results of suggestion system showed that the forecasting system better the ARIMA model than the exponential smoothing method. It has been assumed that the plan for power supply depending on demand and system operation could be performed efficiently by means of using such power load forecasting.