Although separation of touching pigs in real-time is an important issue for a 24-h pig monitoring system, it is challenging to separate accurately the touching pigs in a crowded pig room. In this study, we propose a separation method for touching pigs using the information generated from Convolutional Neural Network(CNN). Especially, we apply one of the CNN-based object detection methods(i.e., You Look Only Once, YOLO) to solve the touching objects separation problem in an active manner. First, we evaluate and select the bounding boxes generated from YOLO, and then separate touching pigs by analyzing the relations between the selected bounding boxes. Our experimental results show that the proposed method is more effective than widely-used methods for separating touching pigs, in terms of both accuracy and execution time.

Maximum intensity projection (MIP) is a common visualization technique in medical imaging system. A typical method to improve the performance of MIP is empty space leaping, which skips unnecessary area. This research proposes a new method to improve the existing empty space leaping. In order to skip more regions, we introduce a variety of acceleration strategies that use some tolerance given by the user to take part in image quality loss. Each proposed method shows various image quality and speed, and this study compares them to select the best one. Experimental results show that it is most efficient to add a constant tolerance function when the image quality required by the user is low. Conversely, when the user required image quality is high, a function with a low tolerance of volume center is most effective. Applying the proposed method to general MIP visualization can generate a relatively high quality image in a short time.

In this paper, we propose a method to classify the skeletal maturity with a small amount of hand wrist X-ray image using deep learning-based meta-learning. General deep-learning techniques require large amounts of data, but in many cases, these data sets are not available for practical application. Lack of learning data is usually solved through transfer learning using pre-trained models with large data sets. However, transfer learning performance may be degraded due to over fitting for unknown new task with small data, which results in poor generalization capability. In addition, medical images require high cost resources such as a professional manpower and mcuh time to obtain labeled data. Therefore, in this paper, we use meta-learning that can classify using only a small amount of new data by pre-trained models trained with various learning tasks. First, we train the meta-model by using a separate data set composed of various learning tasks. The network learns to classify the bone maturity using the bone maturity data composed of the radiographs of the wrist. Then, we compare the results of the classification using the conventional learning algorithm with the results of the meta learning by the same number of learning data sets.

In this paper, we propose an image enhancement method based on Multi-Scale Retinex theory that designs Unsharp Masking Filter (UMF) and emphasizes the contrast ratio adaptively. Unsharp Masking (UM) technique emphasizes image sharpness and improves contrast ratio by adding high frequency component to the original image. The high frequency component is obtained by differentiating between original image and low frequency image. In this paper, we present how to design an UMF kernel and to adaptively apply it to increase the contrast ratio according to multi-scale retinex theory which resembles human visual system. Experimental results show that the proposed method has better quantitative performance indexes such as PSNR, ambe & SSIM and better qualitative feature like halo artifact suppression.

In this paper, we suggest a new system that detects abnormal region of skim. First, an illumination elimination algorithm which uses LAB color model is processed on input facial image to obtain robust facial image for illumination, and then gabor filter is processed to detect the reactivity of discontinuity. And last, the density-based spatial clustering of applications with noise(DBSCAN) algorithm is processed to classify areas of wrinkles, dots, and other skin diseases. This method allows the user to check the skin condition of the images taken in real life.

Recently, human life is getting longer due to change of living environment and development of medical technology, and silver medical technology has been in the limelight. Geriatric skin disease is difficult to detect early, and when it is missed, it becomes a malignant disease and is difficult to treatment. Melanoma is one of the most common diseases of geriatric skin disease and initially has a similar modality with the nevus. In order to overcome this problem, we attempted to perform a feature analysis in order to attempt automatic detection of melanoma-like lesions. In this paper, one is first order analysis using information of pixels in radiomic feature. The other is a gray-level co-occurrence matrix and a gray level run length matrix, which are feature extraction methods for converting image information into a matrix. The features were extracted through these analyses. And classification is implemented by SVM.

In this paper, we describe an algorithm for extracting depth information from a single image based on CNN. When acquiring three-dimensional information from a single two-dimensional image using a deep-learning technique, it is difficult to accurately predict the edge portion of the depth image because it is a part where the depth changes abruptly. in this paper, we introduce the concept of pixel-wise confidence to take advantage of these characteristics. We propose an algorithm that estimates depth information from a highly reliable flat part and propagates it to the edge part to improve the accuracy of depth estimation.

Alzheimer's disease has the significant factors for the both specific and characteristic features according to the disease progressing that are the volumetry and surface area by the brain hippocampus shrinking and thinning. However, we have suggested a shape analysis to calculate the variance by the roughness, coarseness or uneven surface on 3D MR images. For the reasons we have presented two methods: the first method is the distance calculation from major axis to edge points and the second method is the distance calculation from centroidal point to edge points on a coronal plane. Then we selected the shortest distance and the longest distance in each slice and analyzed the ANOVA and average distances. Consequently we obtained the available and great results by the longest distance of the axial and centroidal point. The results of average distances were 44.85(AD), 45.04(MCI) and 49.31(NC) from the axial points and 39.30(AD), 39.58(MCI) and 44.78(NC) from centroidal points respectively. Finally the distance variations for the easily recognized visualization were shown by the color mapping. This research could be provided an indicator of biomarkers that make diagnosis and prognosis the Alzheimer's diseases in the future.

Since a large amount of multimedia image has dramatically increased, it is important to search semantically relevant image. Thus, several semantic image annotation methods using RDF(Resource Description Framework) model in mobile environment are introduced. Earlier studies on annotating image semantically focused on both the image tag and the context-aware information such as temporal and spatial data. However, in order to fully express their semantics of image, we need more annotations which are described in RDF model. In this paper, we propose an annotation method inferencing with RDFS entailment rules and user defined rules. Our approach implemented in Moment system shows that it can more fully represent the semantics of image with more annotation triples.

The distraction of the driver's attention causes as much traffic accidents as drowsiness driving. Yet though there have been many studies on drowsiness driving, research on distraction driving is insufficient. In this paper, we divide distraction of attention into visual distraction and cognitive distraction and analyze the EEG of subjects while viewing images of distracting situations. The results show that more information is received and processed when distractions occur. It is confirmed that the probability of accident increases when the driver receives overwhelming amount of information that he or she cannot concentrate on driving.

This paper proposes a method to improve the performance of face recognition via super-resolution method using sparse representation and deep learning from low-resolution facial images. Recently, there have been many researches on ultra-high-resolution images using deep learning techniques, but studies are still under way in real-time face recognition. In this paper, we combine the sparse representation and deep learning to generate super-resolution images to improve the performance of face recognition. We have also improved the processing speed by designing in parallel structure when applying sparse representation. Finally, experimental results show that the proposed method is superior to conventional methods on various images.

Line-laser beams are used for depth measurement of welding beads along the circumference of a pipe. For this, first we project a line-laser beam on an rotating pipe and take a sequence of images of the beam projected on the pipe using a CCD camera. Second, the projected line laser beam in each image is detected, converted into a thin curve. Finally measure the distance between the thinned curve and an imaginary line. When a line-laser beam is projected to a rough metal surface such as arc welding beads, the beam is severely scattered. This severe scattering makes the thinned curve perturbed. In this paper, we propose a thinning method robust against scattering of line lasers. First, we extract a projected line laser beam region using an adaptive threshold. Second, we model a thinned curve with a spline curve with control points. Next, we adjust the control points to fit the curve to the projected line-laser beam. Finally, we take a weighted mean of thin curves on a sequence of image frames. Experiments shows that the proposed thinning method results in a thinning curve, which is smooth and fit to the projected line-laser beam with small error.

With recent development of high-speed wide-area wireless networks and wide spread of highperformance wireless devices, the demand on seamless video streaming services in Long Term Evolution (LTE) network environments is ever increasing. To meet the demand and provide enhanced Quality of Experience (QoE) with mobile users, the Dynamic Adaptive Streaming over HTTP (DASH) has been actively studied to achieve QoE enhanced video streaming service in dynamic network environments. However, the existing DASH algorithm to select the quality of requesting video segments is based on a procedural algorithm so that it reveals a limitation to adapt its performance to dynamic network situations. To overcome this limitation this paper proposes a novel quality selection mechanism based on a Deep Q-Network (DQN) model, the DQN-based DASH ABR($DQN_{ABR}$) mechanism. The $DQN_{ABR}$ mechanism replaces the existing DASH ABR algorithm with an intelligent deep learning model which optimizes service quality to mobile users through reinforcement learning. Compared to the existing approaches, the experimental analysis shows that the proposed solution outperforms in terms of adapting to dynamic wireless network situations and improving QoE experience of end users.

Recently, services provided to consumers are increasingly being combined with big data such as low-priced shopping, customized advertisement, and product recommendation. With the increasing importance of big data, the web crawler that collects data from the web has also become important. However, there are two problems with existing web crawlers. First, if the URL is hidden from the link, it can not be accessed by the URL. The second is the inefficiency of fetching more data than the user wants. Therefore, in this paper, through the Casper.js which can control the DOM in the headless brwoser, DOM event is generated by accessing the URL to the hidden link. We also propose an intelligent web crawler system that allows users to make steps to fine-tune both Structured and unstructured data to bring only the data they want. Finally, we show the superiority of the proposed crawler system through the performance evaluation results of the existing web crawler and the proposed web crawler.

To understand the meaning of data is a common goal of research on unstructured data. Among these unstructured data, there are difficulties in analyzing the meaning of unstructured data related to corpus and sentences. In the existing researches, the researchers used LSA to select sentences with the most similar meaning to specific words of the sentences. However, it is problematic to examine many sentences continuously. In order to solve unstructured data classification problem, several search sites are available to classify the frequency of words and to serve to users. In this paper, we propose a method of classifying documents by using the frequency of similar words, and the frequency of non-relevant words to be applied as weights, and storing them in terms of a multi-pattern storage. We use Tensorflow's Softmax to the nearby sentences for machine learning, and utilize it for unstructured data analysis and the inference of traffic information.

Recently, the importance of intelligent personalized services is increasing in the era of the 4th Industrial Revolution. In this paper, we proposes design and implementation of the agent recruitment service system for the intelligent headhunting service. The proposed service system has designed and implemented an intelligent service system to efficiently recruit and manage applicants(job seekers). In addition, we develops the modules that can analyze the profile information of the applicant(job seeker), and implement the algorithm to predict the talent verification and analyze the talent type. The proposed service system is expected to contribute to designing the innovative and new service models for the intelligent agent recruitment service system.

Operational Transformation (OT) algorithms for real-time collaborative editing systems are becoming increasingly important due to the increased demand for collaborative data processing. The operational transformation algorithm is a technique for real-time concurrency control and consistency maintenance with non-locking technique, and many studies have been conducted to overcome three issues of convergence, causality-prevention, and intention-prevention. However, previous work has the disadvantage of wasting memory by storing all operations that occurred during an edit operation in the history buffer to solve this problem. Therefore, we propose a memory-saving real-time collaborative editing system that maintains a constant memory space and concurrency control through a method of applying the valid-time to each user-generated operation in order to reduce memory waste. This system prevents long-term memory occupation of client-generated operations, thus it reduces the space and time complexity even with low-rate of collaboration work, so that the performance degradation avoids.

In this paper, the customized sports convergence contents curation system is proposed for activation of life sports. The proposed system collects and analyzes profile of social sports group (club, society, etc.) for recommending optimized sports convergence contents to user. In addition, the feedback based on the recommendation result from the user is continuously reflected and the optimal recommendation is made possible. For the system evaluation, the proposed system is tested to 300 users (about 20 sports team) for about 3 months and the system is verified by analyzing the initial recommendation results and recommendation results reflected by user feedback.

In order to control the quadcopter using reinforcement learning, hovering of 1-axial drones prototype is implemented through reinforcement learning. A complementary filter is used to measure the correct angle, and the range of angles is from -180 degrees to +180 degrees using modified complementary filter. The policy gradient method is used together with the REINFORCE algorithm for reinforcement learning. The prototype learned in this way confirmed the difference in performance depending on the length of the episode.

Hearing-impaired people are increasing rapidly due to the global aging trend. Early detection of hearing loss requires an easy-to-use audiometry device for the public. Existing audiometry systems were developed as PC-based, PDA-based, or smartphone apps. These devices were often dependent on specific software platforms and hardware platforms. In this paper, we tried to improve software platform compatibility by using cross platform, and tried to implement IoT-based pure tone audiometry device which does not require sound pressure level correction due to hardware differences. Pure tone audiometry is available in a variety of ways depending on the type of hearing loss and age. Using the IoT-based audiometry device implemented in this paper, it will be possible for an app developer who lacks hardware knowledge to easily develop an app with various scenarios for hearing screening. The results of this study will contribute to overcoming the software and hardware dependency in the development of IoT-based healthcare device.

Recently, big data has been growing rapidly due to the development of IT technology. Especially in the medical field, big data is utilized to provide services such as patient-customized medical care, disease management and disease prediction. In Korea, 'National Health Alarm Service' is provided by National Health Insurance Corporation. However, the prediction model has a problem of short-term prediction within 3 days and unreliability of social data used in prediction model. In order to solve these problems, this paper proposes a disease prediction model using medicine prescription data generated from actual patients. This model predicts the total number of patients and the risk of disease in each region and uses the ARIMA model for long-term predictions.

Crowdfunding has seen an enormous rise, becoming a new alternative funding source for emerging startup companies in recent years. Despite the huge success of crowdfunding, it has been reported that only around 40% of crowdfunding campaigns successfully raise the desired goal amount. The purpose of this study is to investigate key factors influencing successful fundraising on crowdfunding platforms. To this end, we mainly focus on contents of project campaigns, particularly their linguistic cues as well as multiple features extracted from project information and multimedia contents. We reveal which of these features are useful for predicting success of crowdfunding campaigns, and then build a predictive model based on those selected features. Our experimental results demonstrate that the built model predicts the success or failure of a crowdfunding campaign with 86.15% accuracy.

The 360 degree panoramas are picture of a wide range of images on one screen, so we can see a fairly wide range at a time. In particular, cylinderical panoramas are the most widely used spherical image, and its left and right viewing angles reach 360 degree, so you can observe front, rear, left, and right at once. Using 360 degree panorama, all directions can be monitored at the same time, so all directions can be effectively monitored compared to other methods. In this paper, we develop a system to recognize and track the movement of moving objects on a 360 degree panorama, and then present and verify the experimental results. For this goals, first, we developed a system to recognize moving objects in 360 degree panorama using DoF(Difference of Frame) algorithm. Second, based on the TLD algorithm, we developed an application that can track a specific single moving object in a 360 degree panorama and presented the experimental results.

Detection of vanishing point is a challenging task in the situations where there are several structures with straight lines. Commonly used approaches for determining vanishing points involves finding the straight lines using edge detection and Hough transform methods. This approach often fails to perform effectively when there are a lot of straight lines found. The lines not meeting at a vanishing point are considered to be noises. In such situation, finding right candidate lines for detecting vanishing points is not a simple task. This paper proposes to use reference objects for vanishing point detection. By analyzing a reference object, it identifies the contour of the object, and derives a polygon from the contour information. Then the edges of the detected polygon are used to find the vanishing points. Our experimental results show that the proposed approach can detect vanishing points with comparable accuracy to the existing edge detection based method. Our approach can also be applied effectively even to complex situations, where too many lines generated by the existing methods make it difficult to select right lines for the vanishing points.

Documentation is a typical method that helps students to understand a program for implementation and execute error correction and maintenance cost-effectively. Guidelines for components that organize documentation should be provided to enable to express computational thinking and such components also should be linked to coding process. In this paper, we focused on the documentation format to guide elementary school students, who were beginners in computational thinking, to express computational thinking. The improvement in the expression of computational thinking was analyzed based on the documentation format applied to the class, and practical tips on the importance of components organizing the documentation format were proposed.

Today, the life cycle of mobile games is getting shorter. In order to overcome this problem, we tried to summarize the strategies and marketing types of game marketing. Recently, I considered the TV advertisement that used the star of the marketing form of mobile game. In this research, user survey was analyzed based on correlation between star marketing and mobile game TV advertisement. And, through an interview with Delphi experts, I summarized the important elements of TV advertisement for mobile games. In conclusion, this research presented a total of five elements on a successful mobile game marketing strategy.The direction of mobile game advertising can promote a successful game that needs to set effective advertising marketing strategy, based on the basic nature of the game and main target.