• Journal of the Institute of Electronics and Information Engineers
• /
• v.51 no.4
• /
• pp.144-159
• /
• 2014

Techniques for making a single panoramic image using multiple pictures are widely studied in many areas such as computer vision, computer graphics, etc. The panorama image can be applied to various fields like virtual reality, robot vision areas which require wide-angled shots as an useful way to overcome the limitations such as picture-angle, resolutions, and internal informations of an image taken from a single camera. It is so much meaningful in a point that a panoramic image usually provides better immersion feeling than a plain image. Although there are many ways to build a panoramic image, most of them are using the way of extracting feature points and matching points of each images for making a single panoramic image. In addition, those methods use the RANSAC(RANdom SAmple Consensus) algorithm with matching points and the Homography matrix to transform the image. The SURF(Speeded Up Robust Features) algorithm which is used in this paper to extract featuring points uses an image's black and white informations and local spatial informations. The SURF is widely being used since it is very much robust at detecting image's size, view-point changes, and additionally, faster than the SIFT(Scale Invariant Features Transform) algorithm. The SURF has a shortcoming of making an error which results in decreasing the RANSAC algorithm's performance speed when extracting image's feature points. As a result, this may increase the CPU usage occupation rate. The error of detecting matching points may role as a critical reason for disqualifying panoramic image's accuracy and lucidity. In this paper, in order to minimize errors of extracting matching points, we used $3{\times}3$ region's RGB pixel values around the matching points' coordinates to perform intermediate filtering process for removing wrong matching points. We have also presented analysis and evaluation results relating to enhanced working speed for producing a panorama image, CPU usage rate, extracted matching points' decreasing rate and accuracy.

• Journal of the Korean Society of Surveying, Geodesy, Photogrammetry and Cartography
• /
• v.37 no.3
• /
• pp.199-208
• /
• 2019

As the number of available satellites increases and technology advances, image information outputs are becoming increasingly diverse and a large amount of data is accumulating. In this study, we propose a change detection method for high-resolution satellite images that uses transfer learning and a deep learning network to overcome the limit caused by insufficient training data via the use of pre-trained information. The deep learning network used in this study comprises convolutional layers to extract the spatial and spectral information and convolutional long-short term memory layers to analyze the time series information. To use the learned information, the two initial convolutional layers of the change detection network are designed to use learned values from 40,000 patches of the ISPRS (International Society for Photogrammertry and Remote Sensing) dataset as initial values. In addition, 2D (2-Dimensional) and 3D (3-dimensional) kernels were used to find the optimized structure for the high-resolution satellite images. The experimental results for the KOMPSAT-3A (KOrean Multi-Purpose SATllite-3A) satellite images show that this change detection method can effectively extract changed/unchanged pixels but is less sensitive to changes due to shadow and relief displacements. In addition, the change detection accuracy of two sites was improved by using 3D kernels. This is because a 3D kernel can consider not only the spatial information but also the spectral information. This study indicates that we can effectively detect changes in high-resolution satellite images using the constructed image information and deep learning network. In future work, a pre-trained change detection network will be applied to newly obtained images to extend the scope of the application.

• Journal of Intelligence and Information Systems
• /
• v.24 no.3
• /
• pp.1-19
• /
• 2018

Large amount of data is now available for research and business sectors to extract knowledge from it. This data can be in the form of unstructured data such as audio, text, and image data and can be analyzed by deep learning methodology. Deep learning is now widely used for various estimation, classification, and prediction problems. Especially, fashion business adopts deep learning techniques for apparel recognition, apparel search and retrieval engine, and automatic product recommendation. The core model of these applications is the image classification using Convolutional Neural Networks (CNN). CNN is made up of neurons which learn parameters such as weights while inputs come through and reach outputs. CNN has layer structure which is best suited for image classification as it is comprised of convolutional layer for generating feature maps, pooling layer for reducing the dimensionality of feature maps, and fully-connected layer for classifying the extracted features. However, most of the classification models have been trained using online product image, which is taken under controlled situation such as apparel image itself or professional model wearing apparel. This image may not be an effective way to train the classification model considering the situation when one might want to classify street fashion image or walking image, which is taken in uncontrolled situation and involves people's movement and unexpected pose. Therefore, we propose to train the model with runway apparel image dataset which captures mobility. This will allow the classification model to be trained with far more variable data and enhance the adaptation with diverse query image. To achieve both convergence and generalization of the model, we apply Transfer Learning on our training network. As Transfer Learning in CNN is composed of pre-training and fine-tuning stages, we divide the training step into two. First, we pre-train our architecture with large-scale dataset, ImageNet dataset, which consists of 1.2 million images with 1000 categories including animals, plants, activities, materials, instrumentations, scenes, and foods. We use GoogLeNet for our main architecture as it has achieved great accuracy with efficiency in ImageNet Large Scale Visual Recognition Challenge (ILSVRC). Second, we fine-tune the network with our own runway image dataset. For the runway image dataset, we could not find any previously and publicly made dataset, so we collect the dataset from Google Image Search attaining 2426 images of 32 major fashion brands including Anna Molinari, Balenciaga, Balmain, Brioni, Burberry, Celine, Chanel, Chloe, Christian Dior, Cividini, Dolce and Gabbana, Emilio Pucci, Ermenegildo, Fendi, Giuliana Teso, Gucci, Issey Miyake, Kenzo, Leonard, Louis Vuitton, Marc Jacobs, Marni, Max Mara, Missoni, Moschino, Ralph Lauren, Roberto Cavalli, Sonia Rykiel, Stella McCartney, Valentino, Versace, and Yve Saint Laurent. We perform 10-folded experiments to consider the random generation of training data, and our proposed model has achieved accuracy of 67.2% on final test. Our research suggests several advantages over previous related studies as to our best knowledge, there haven't been any previous studies which trained the network for apparel image classification based on runway image dataset. We suggest the idea of training model with image capturing all the possible postures, which is denoted as mobility, by using our own runway apparel image dataset. Moreover, by applying Transfer Learning and using checkpoint and parameters provided by Tensorflow Slim, we could save time spent on training the classification model as taking 6 minutes per experiment to train the classifier. This model can be used in many business applications where the query image can be runway image, product image, or street fashion image. To be specific, runway query image can be used for mobile application service during fashion week to facilitate brand search, street style query image can be classified during fashion editorial task to classify and label the brand or style, and website query image can be processed by e-commerce multi-complex service providing item information or recommending similar item.

• The Transactions of the Korea Information Processing Society
• /
• v.4 no.12
• /
• pp.3159-3174
• /
• 1997

This paper reports a research to develop a methodology and a tool for understanding of very large and complex real-time software. The methodology and the tool mostly developed by the author are called the Architecture-based Real-time Software Understanding (ARSU) and the Software Re/reverse-engineering Environment (SRE) respectively. Due to size and complexity, it is commonly very hard to understand the software during reengineering process. However the research facilitates scalable re/reverse-engineering of such real-time software based on the architecture of the software in three-dimensional perspectives: structural, functional, and behavioral views. Firstly, the structural view reveals the overall architecture, specification (outline), and the algorithm (detail) views of the software, based on hierarchically organized parent-chi1d relationship. The basic building block of the architecture is a software Unit (SWU), generated by user-defined criteria. The architecture facilitates navigation of the software in top-down or bottom-up way. It captures the specification and algorithm views at different levels of abstraction. It also shows the functional and the behavioral information at these levels. Secondly, the functional view includes graphs of data/control flow, input/output, definition/use, variable/reference, etc. Each feature of the view contains different kind of functionality of the software. Thirdly, the behavioral view includes state diagrams, interleaved event lists, etc. This view shows the dynamic properties or the software at runtime. Beside these views, there are a number of other documents: capabilities, interfaces, comments, code, etc. One of the most powerful characteristics of this approach is the capability of abstracting and exploding these dimensional information in the architecture through navigation. These capabilities establish the foundation for scalable and modular understanding of the software. This approach allows engineers to extract reusable components from the software during reengineering process.

• Journal of the Institute of Electronics Engineers of Korea TC
• /
• v.49 no.1
• /
• pp.31-46
• /
• 2012

The most critical technology to implement ubiquitous health care is Ubiquitous Sensor Network (USN) technology which makes use of various sensor technologies, processor integration technology, and wireless network technology-Radio Frequency Identification (RFID) and Wireless Sensor Network (WSN)-to easily gather and monitor actual physical environment information from a remote site. With the feature, the USN technology can make the information technology of the existing virtual space expanded to actual environments. However, although the RFID and the WSN have technical similarities and mutual effects, they have been recognized to be studied separately, and sufficient studies have not been conducted on the technical integration of the RFID and the WSN. Therefore, EPCglobal which realized the issue proposed the EPC Sensor Network to efficiently integrate and interoperate the RFID and WSN technologies based on the international standard EPCglobal network. The proposed EPC Sensor Network technology uses the Complex Event Processing method in the middleware to integrate data occurring through the RFID and the WSN in a single environment and to interoperate the events based on the EPCglobal network. However, as the EPC Sensor Network technology continuously performs its operation even in the case that the minimum conditions are not to be met to find complex events in the middleware, its operation cost rises. Moreover, since the technology is based on the EPCglobal network, it can neither perform its operation only for the sake of sensor data, nor connect or interoperate with each information system in which the most important information in the ubiquitous computing environment is saved. Therefore, to address the problems of the existing system, we proposed the design and implementation of USN integration management system. For this, we first proposed an integration system that manages RFID and WSN data based on Session Initiation Protocol (SIP). Secondly, we defined the minimum conditions of the complex events to detect unnecessary complex events in the middleware, and proposed an algorithm that can extract complex events only when the minimum conditions are to be met. To evaluate the performance of the proposed methods we implemented SIP-based integration management system.

• Journal of Intelligence and Information Systems
• /
• v.18 no.1
• /
• pp.39-57
• /
• 2012

Recently, due to the introduction of high-tech equipment in interactive exhibits, many people's attention has been concentrated on Interactive exhibits that can double the exhibition effect through the interaction with the audience. In addition, it is also possible to measure a variety of audience reaction in the interactive exhibition. Among various audience reactions, this research uses the change of the facial features that can be collected in an interactive exhibition space. This research develops an artificial neural network-based prediction model to predict the response of the audience by measuring the change of the facial features when the audience is given stimulation from the non-excited state. To present the emotion state of the audience, this research uses a Valence-Arousal model. So, this research suggests an overall framework composed of the following six steps. The first step is a step of collecting data for modeling. The data was collected from people participated in the 2012 Seoul DMC Culture Open, and the collected data was used for the experiments. The second step extracts 64 facial features from the collected data and compensates the facial feature values. The third step generates independent and dependent variables of an artificial neural network model. The fourth step extracts the independent variable that affects the dependent variable using the statistical technique. The fifth step builds an artificial neural network model and performs a learning process using train set and test set. Finally the last sixth step is to validate the prediction performance of artificial neural network model using the validation data set. The proposed model is compared with statistical predictive model to see whether it had better performance or not. As a result, although the data set in this experiment had much noise, the proposed model showed better results when the model was compared with multiple regression analysis model. If the prediction model of audience reaction was used in the real exhibition, it will be able to provide countermeasures and services appropriate to the audience's reaction viewing the exhibits. Specifically, if the arousal of audience about Exhibits is low, Action to increase arousal of the audience will be taken. For instance, we recommend the audience another preferred contents or using a light or sound to focus on these exhibits. In other words, when planning future exhibitions, planning the exhibition to satisfy various audience preferences would be possible. And it is expected to foster a personalized environment to concentrate on the exhibits. But, the proposed model in this research still shows the low prediction accuracy. The cause is in some parts as follows : First, the data covers diverse visitors of real exhibitions, so it was difficult to control the optimized experimental environment. So, the collected data has much noise, and it would results a lower accuracy. In further research, the data collection will be conducted in a more optimized experimental environment. The further research to increase the accuracy of the predictions of the model will be conducted. Second, using changes of facial expression only is thought to be not enough to extract audience emotions. If facial expression is combined with other responses, such as the sound, audience behavior, it would result a better result.

• Journal of Intelligence and Information Systems
• /
• v.19 no.2
• /
• pp.1-20
• /
• 2013

Nowadays, social network is a huge communication platform for providing people to connect with one another and to bring users together to share common interests, experiences, and their daily activities. Users spend hours per day in maintaining personal information and interacting with other people via posting, commenting, messaging, games, social events, and applications. Due to the growth of user's distributed information in social network, there is a great potential to utilize the social data to enhance the quality of recommender system. There are some researches focusing on social network analysis that investigate how social network can be used in recommendation domain. Among these researches, we are interested in taking advantages of the interaction between a user and others in social network that can be determined and known as social relationship. Furthermore, mostly user's decisions before purchasing some products depend on suggestion of people who have either the same preferences or closer relationship. For this reason, we believe that user's relationship in social network can provide an effective way to increase the quality in prediction user's interests of recommender system. Therefore, social relationship between users encountered from social network is a common factor to improve the way of predicting user's preferences in the conventional approach. Recommender system is dramatically increasing in popularity and currently being used by many e-commerce sites such as Amazon.com, Last.fm, eBay.com, etc. Collaborative filtering (CF) method is one of the essential and powerful techniques in recommender system for suggesting the appropriate items to user by learning user's preferences. CF method focuses on user data and generates automatic prediction about user's interests by gathering information from users who share similar background and preferences. Specifically, the intension of CF method is to find users who have similar preferences and to suggest target user items that were mostly preferred by those nearest neighbor users. There are two basic units that need to be considered by CF method, the user and the item. Each user needs to provide his rating value on items i.e. movies, products, books, etc to indicate their interests on those items. In addition, CF uses the user-rating matrix to find a group of users who have similar rating with target user. Then, it predicts unknown rating value for items that target user has not rated. Currently, CF has been successfully implemented in both information filtering and e-commerce applications. However, it remains some important challenges such as cold start, data sparsity, and scalability reflected on quality and accuracy of prediction. In order to overcome these challenges, many researchers have proposed various kinds of CF method such as hybrid CF, trust-based CF, social network-based CF, etc. In the purpose of improving the recommendation performance and prediction accuracy of standard CF, in this paper we propose a method which integrates traditional CF technique with social relationship between users discovered from user's behavior in social network i.e. Facebook. We identify user's relationship from behavior of user such as posts and comments interacted with friends in Facebook. We believe that social relationship implicitly inferred from user's behavior can be likely applied to compensate the limitation of conventional approach. Therefore, we extract posts and comments of each user by using Facebook Graph API and calculate feature score among each term to obtain feature vector for computing similarity of user. Then, we combine the result with similarity value computed using traditional CF technique. Finally, our system provides a list of recommended items according to neighbor users who have the biggest total similarity value to the target user. In order to verify and evaluate our proposed method we have performed an experiment on data collected from our Movies Rating System. Prediction accuracy evaluation is conducted to demonstrate how much our algorithm gives the correctness of recommendation to user in terms of MAE. Then, the evaluation of performance is made to show the effectiveness of our method in terms of precision, recall, and F1-measure. Evaluation on coverage is also included in our experiment to see the ability of generating recommendation. The experimental results show that our proposed method outperform and more accurate in suggesting items to users with better performance. The effectiveness of user's behavior in social network particularly shows the significant improvement by up to 6% on recommendation accuracy. Moreover, experiment of recommendation performance shows that incorporating social relationship observed from user's behavior into CF is beneficial and useful to generate recommendation with 7% improvement of performance compared with benchmark methods. Finally, we confirm that interaction between users in social network is able to enhance the accuracy and give better recommendation in conventional approach.

• Proceedings of the Safety Management and Science Conference
• /
• 2004.11a
• /
• pp.355-380
• /
• 2004

I test the hypothesis that the gradual diffusion of information across asset markets leads to cross-asset return predictability in Korea. Using thirty-six industry portfolios and the broad market index as our test assets, I establish several key results. First, a number of industries such as semiconductor, electronics, metal, and petroleum lead the stock market by up to one month. In contrast, the market, which is widely followed, only leads a few industries. Importantly, an industry's ability to lead the market is correlated with its propensity to forecast various indicators of economic activity such as industrial production growth. Consistent with our hypothesis, these findings indicate that the market reacts with a delay to information in industry returns about its fundamentals because information diffuses only gradually across asset markets. Traditional theories of asset pricing assume that investors have unlimited information-processing capacity. However, this assumption does not hold for many traders, even the most sophisticated ones. Many economists recognize that investors are better characterized as being only boundedly rational(see Shiller(2000), Sims(2201)). Even from casual observation, few traders can pay attention to all sources of information much less understand their impact on the prices of assets that they trade. Indeed, a large literature in psychology documents the extent to which even attention is a precious cognitive resource(see, eg., Kahneman(1973), Nisbett and Ross(1980), Fiske and Taylor(1991)). A number of papers have explored the implications of limited information- processing capacity for asset prices. I will review this literature in Section II. For instance, Merton(1987) develops a static model of multiple stocks in which investors only have information about a limited number of stocks and only trade those that they have information about. Related models of limited market participation include brennan(1975) and Allen and Gale(1994). As a result, stocks that are less recognized by investors have a smaller investor base(neglected stocks) and trade at a greater discount because of limited risk sharing. More recently, Hong and Stein(1999) develop a dynamic model of a single asset in which information gradually diffuses across the investment public and investors are unable to perform the rational expectations trick of extracting information from prices. Hong and Stein(1999). My hypothesis is that the gradual diffusion of information across asset markets leads to cross-asset return predictability. This hypothesis relies on two key assumptions. The first is that valuable information that originates in one asset reaches investors in other markets only with a lag, i.e. news travels slowly across markets. The second assumption is that because of limited information-processing capacity, many (though not necessarily all) investors may not pay attention or be able to extract the information from the asset prices of markets that they do not participate in. These two assumptions taken together leads to cross-asset return predictability. My hypothesis would appear to be a very plausible one for a few reasons. To begin with, as pointed out by Merton(1987) and the subsequent literature on segmented markets and limited market participation, few investors trade all assets. Put another way, limited participation is a pervasive feature of financial markets. Indeed, even among equity money managers, there is specialization along industries such as sector or market timing funds. Some reasons for this limited market participation include tax, regulatory or liquidity constraints. More plausibly, investors have to specialize because they have their hands full trying to understand the markets that they do participate in

• Tuberculosis and Respiratory Diseases
• /
• v.58 no.6
• /
• pp.590-599
• /
• 2005

Object : Cigarette smoking is a major cause of mucus hypersecretion, which is a pathophysiological feature of many inflammatory airway diseases. Mucins, which are an important part of the airway mucus, are synthesized from the Muc gene in airway epithelial cells. However, the signaling pathways for cigarette smoke-induced mucin synthesis are unknown. The aim of this study was to determine the signal pathway for smoking induced Muc5ac gene expression. Methods : A549 cells were cultured and transiently transfected with the Muc5ac promoter fragment. These cells were stimulated with 5% cigarette smoke extract (CSE) alone or with CSE after a pretreatment with various signal transduction pathway inhibitors (AG1478, PD98059 and SB203580). The Muc5ac promoter activity was examined using the luciferase reporter system, and the level of phosphorylated EGFR, ERK1/2, p38 MAPK and JNK were all examined using Western blot analysis. Muc5ac mRNA expression was also examined using reverse transcriptase polymerase chain reactions (RT-PCR). Results : 1. The peak level of luciferase activity of the Muc5ac promoter was observed at 5% concentration and after 3 hours of incubation with the CSE. The level of EGFR phosphorylation and the luciferase activity of the transfected cells caused by the CSE were significantly suppressed by AG1478 or PD98059 (P<0.01). 2. CSE phosphorylated ERK1/2 or p38 MAPK but not JNK. The Muc5ac mRNA expression level was increased by the CSE but that was suppressed by PD98059 or AG1478. 3. The CSE-induced phosphorylation of ERK1/2 was blocked by PD98059 and that of p38 MAPK was blocked by either PD98059 or SB203580. Either PD98059 or SB203580 suppressed the luciferase activity of the transfected cells (P<0.0001). Conclusion : The Muc5ac mRNA expression level was increased by the CSE. The increased CSE-induced transcriptional activity was mediated via EGF receptor activation, which led to ERK1/2 and p38 MAPK phosphorylation.