• Journal of Computing Science and Engineering
• /
• 제5권3호
• /
• pp.161-166
• /
• 2011

We present a fast algorithm for Korean text extraction and segmentation from subway signboards using smart phone sensors in order to minimize computational time and memory usage. The algorithm can be used as preprocessing steps for optical character recognition (OCR): binarization, text location, and segmentation. An image of a signboard captured by smart phone camera while holding smart phone by an arbitrary angle is rotated by the detected angle, as if the image was taken by holding a smart phone horizontally. Binarization is only performed once on the subset of connected components instead of the whole image area, resulting in a large reduction in computational time. Text location is guided by user's marker-line placed over the region of interest in binarized image via smart phone touch screen. Then, text segmentation utilizes the data of connected components received in the binarization step, and cuts the string into individual images for designated characters. The resulting data could be used as OCR input, hence solving the most difficult part of OCR on text area included in natural scene images. The experimental results showed that the binarization algorithm of our method is 3.5 and 3.7 times faster than Niblack and Sauvola adaptive-thresholding algorithms, respectively. In addition, our method achieved better quality than other methods.

• International journal of advanced smart convergence
• /
• 제9권3호
• /
• pp.260-271
• /
• 2020

Smartphones provide blind users with screenreader as an accessibility tool. However, blind users often experience difficulties accessing online shopping malls via smartphones due to their inconsistent and image-based layouts. To enable screenreader users to get access to the detailed information about products while they are shopping online, we have developed BarrierFreeShop, an accessible mobile shopping application for people with visual impairments. BarrierFreeShop has three accessibility features: (1) layout automation, (2) review summarization, and (3) optical character recognition. We conducted a user study with 80 participants with visual impairments where they were asked to use BarrierFreeShop for a month. The findings revealed the effectiveness of our app in terms of speed and post interview feedback. We have also discovered typical shopping experiences that participants had during the test. This research suggests that computer vision technologies can improve accessibility issues in online shopping malls. In addition, we have confirmed that extracting contents from images help people with visual impairments to get better access to product information.

• IEIE Transactions on Smart Processing and Computing
• /
• 제5권5호
• /
• pp.327-330
• /
• 2016

Currently, the world's elderly population continues to grow at a dramatic rate. As the number of senior citizens increases, detection of someone falling has attracted increasing attention for visual surveillance systems. This paper presents a novel fall-detection algorithm using motion estimation and an integrated spatiotemporal energy map of the object region. The proposed method first extracts a human region using a background subtraction method. Next, we applied an optical flow algorithm to estimate motion vectors, and an energy map is generated by accumulating the detected human region for a certain period of time. We can then detect a fall using k-nearest neighbor (kNN) classification with the previously estimated motion information and energy map. The experimental results show that the proposed algorithm can effectively detect someone falling in any direction, including at an angle parallel to the camera's optical axis.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• 제9권8호
• /
• pp.2948-2963
• /
• 2015

In the field of computer vision, visual surveillance systems have recently become an important research topic. Growth in this area is being driven by both the increase in the availability of inexpensive computing devices and image sensors as well as the general inefficiency of manual surveillance and monitoring. In particular, the ultimate goal for many visual surveillance systems is to provide automatic activity recognition for events at a given site. A higher level of understanding of these activities requires certain lower-level computer vision tasks to be performed. So in this paper, we propose an intelligent activity recognition model that uses a structure learning method and a classification method. The structure learning method is provided as a K2-learning algorithm that generates Bayesian networks of causal relationships between sensors for a given activity. The statistical characteristics of the sensor values and the topological characteristics of the generated graphs are learned for each activity, and then a neural network is designed to classify the current activity according to the features extracted from the multiple sensor values that have been collected. Finally, the proposed method is implemented and tested by using PETS2013 benchmark data.

• 대한임베디드공학회논문지
• /
• 제15권2호
• /
• pp.51-60
• /
• 2020

Optical character recognition (OCR) has been studied for decades because it is very useful in a variety of places. Nowadays, OCR's performance has improved significantly due to outstanding deep learning technology. Thus, there is an increasing demand for commercial-grade but affordable OCR systems. We have developed a low-cost, high-performance OCR system for the industry with the cheapest embedded developer kit that supports GPU acceleration. To achieve high accuracy for industrial use on limited computing resources, we chose a state-of-the-art text recognition algorithm that uses an end-to-end deep learning network as a baseline model. The model was then improved by replacing the feature extraction network with the best one suited to our conditions. Among the various candidate networks, EfficientNet-B3 has shown the best performance: excellent recognition accuracy with relatively low memory consumption. Besides, we have optimized the model written in TensorFlow's Python API using TensorFlow-TensorRT integration and TensorFlow's C++ API, respectively.

• 대한전기학회:학술대회논문집
• /
• 대한전기학회 1994년도 추계학술대회 논문집 학회본부
• /
• pp.379-381
• /
• 1994

Volume holograms recorded in photorefractive materials can find important applications in optical memories and optical computing systems. One problem with a photorefractive hologram is that it gets erased by the readout light. Nondestructive readout can be achieved by hologram fixing, and several fixing methods have been reported. Fixing is accomplished by thermally activated motion of an unknown ionic defect, which neutralizes the electronic space-charge patterns. At room temperature the ionic patterns are stabilized. When the electrons are partially redistributed by light, a net space-charge pattern appears, and tile fixed hologram can be read out. In this paper, theoretical modeling and some experimental results are presented for thermal fixing of volume phase holograms in photorefractive $LiNbO_3$:Fe. Thermal fixing can be done during or after recording and depends on fixing temperature ($100{\sim}200^{\circ}C$ range) and grating length. Fixed Slating can be erased completely at the temperature over $300^{\circ}C$. Theoretical modeling shows weil the compensation of electronic Slating by ionic grating and is in good agreement with experimental results. In experiments the dependence of thermal fixing on temperatures and grating lengths is investigated.

• 한국정보처리학회:학술대회논문집
• /
• 한국정보처리학회 2022년도 추계학술발표대회
• /
• pp.541-543
• /
• 2022

In recent years, retinal disorders have become a serious health concern. Retinal disorders develop slowly and without obvious signs. To avoid vision deterioration, early detection and treatment are critical. Optical coherence tomography (OCT) is a non-invasive and non-contact medical imaging technique used to acquire informative and high-resolution image of retinal area and underlying layers. Disease signs are difficult to detect because OCT images have many areas which are not related to any disease. In this paper, we present a deep learning-based method to perform multi-label classification on a long-tailed OCT dataset. Our method first extracts the region of interest and then performs the classification task. We achieve 98% accuracy, 92% sensitivity, and 99% specificity on our private OCT dataset. Using the heatmap generated from trained convolutional neural network, our method is more robust and explainable than previous approaches because it focuses on areas that contain disease signs.

• 감성과학
• /
• 제12권4호
• /
• pp.393-402
• /
• 2009

본 연구의 목적은 스마트 포토닉 의류와 관련된 다양한 기술 중 광섬유 발광 기능 의류 기술을 기반으로 하여, 스마트 포토닉 스포츠 의류를 위한 모듈화 디자인의 체계적인 방안을 모색하고, 인체 치수를 기반으로 한 광섬유-발광 모듈 조합에 따른 다양한 모듈화 디자인 모형을 제시하는데 있다. 이를 위하여 본 연구에서는 모듈화의 디자인 요건을 살펴보았고, 광섬유-발광 모듈의 크기와 부착 위치를 선정함에 있어서는 모듈화와 인체 치수와의 밀접한 연관성을 고려하여 연구를 진행하였다. 인체 치수와 더불어 컴퓨팅 기기의 적합한 부착 위치와 인체의 동작에 영향을 덜 받는 인체 부위를 종합적으로 고려하여 최종적으로 광섬유-발광 모듈을 부착하기에 적합한 인체상의 부위를 도출하였고, 이를 부위별 실측 치수와 함께 제시하였다. 이를 토대로 스마트 포토닉 의류의 3대 기능 중 스포츠 의류에 가장 필요하면서 적합하다고 판단되는 착용자 안전보호 기능을 중심으로, 스마트 포토닉 기능의 스포츠 의류 상의 재킷(jacket)의 모듈화 모형을 설계하여 제시하였다.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• 제8권7호
• /
• pp.2325-2349
• /
• 2014

Network virtualization technology plays a key role in cloud computing, which serves as an effective approach for provisioning a flexible and highly adaptable shared substrate network to satisfy the demands of various applications or services. Recently, the problem of mapping a virtual network (VN) onto a substrate network has been addressed by various algorithms. However, these algorithms are typically efficient for unicast service-oriented virtual networks, and generally not applicable to multicast service-oriented virtual networks (MVNs). Furthermore, the survivable MVN mapping (SMVNM) problem that considers the survivability of MVN has not been studied and is also the focus of this work. In this research, we discuss SMVNM problem under regional failures in the substrate network and propose an efficient algorithm for solving this problem. We first propose a framework and formulate the SMVNM problem with the objective of minimizing mapping cost by using mixed integer linear programming. Then we design an efficient heuristic to solve this problem and introduce several optimizations to achieve the better mapping solutions. We validate and evaluate our framework and algorithms by conducting extensive simulations on different realistic networks under various scenarios, and by comparing with existing approaches. Our simulation experiments and results show that our approach outperforms existing solutions.

• 한국진공학회:학술대회논문집
• /
• 한국진공학회 2012년도 제42회 동계 정기 학술대회 초록집
• /
• pp.251-251
• /
• 2012

A lot of studies are undergoing on the magneto-optical (MO) properties of patterned magnetic systems for the reason that they have potential application to information technology such as ultrahigh-speed computing. Moreover, they can be considered as the future candidates for high-density MO storage devices. Not only the technical aspects, but there have been also tremendous interests in studying their properties related to the fundamental physics. The MO Kerr-rotation effects (both in reflected and the diffracted modes) and the magnetic force microscopy (MFM) are very useful techniques to investigate the micromagnetic properties of such periodic structures. Hence, in this study, we report on the MO properties of bilayered Cobalt (Co)/ nickel (Ni) micro-patterned anti-dot arrays. Such a ferromagnetic structure was made by sequentially depositing co (40 nm)/Ni (5 nm) bilayer on a Si substrate. The anti-dot patterning with hole diameter of $1{\mu}m$ was done only on the upper Co layer using photolithography technique, while the Ni underlayer was kept uniform. The longitudinal Kerr rotation (LKR) of the zeroth- and the first-order diffracted beams were measured at an incidence of $30^{\circ}$ by using a photoelastic modulator method. The external magnetic field was applied perpendicularly to the reflected and the diffracted beams using an electromagnet capable of a maximum field of ${\pm}5$ kOe. Significantly, it was observed that the LKR of the first-order diffracted beam is nearly 4 times larger than that of the zeroth-order beam. The simulated results for the hysteresis loops matched qualitatively well with the experimentally obtained ones. In conjunction with the LKR, we also investigated the magnetic-domain structure by using a MFM system, which were analyzed to elucidate the origin of the enhanced MO rotation.