Due to the low output rate and time delay of vehicle's navigation results, the electro-optical tracking system(EOTS) cannot estimate accurate target positions. If an inertial measurement unit(IMU) is additionally mounted into the EOTS and inertial navigation system(INS) is constructed, the high navigation output rate can be obtained. And the time-delay can be compensated by using the augmented Kalman filter. An accurate initial attitude is required in order to have accurate navigation outputs. In this paper, an attitude determination algorithm is proposed using the augmented Kalman filter in order to compensate measurement delay of the EOTS and have accurate initial attitude. The proposed initial attitude determination algorithm consists of an augmented Kalman filter, an INS, and an integrated Kalman filter. The augmented Kalman filter compensates the time-delay of the vehicle's navigation results and the integrated Kalman filter estimates the navigation error of the INS. In order to evaluate performance of the proposed algorithm, vehicle's navigation outputs and IMU measurements were generated using sensors' model-based measurement generator and initial attitude estimation errors of the proposed algorithm and the conventional algorithm without the augmented Kalman filter were compared for the generated measurements. The evaluation results show that the proposed algorithm has better accuracy.

Recently, convolutional neural networks (CNNs) have been widely used with excellent performance in various computer vision fields, including super-resolution (SR). However, CNN is computationally intensive and requires a lot of memory, making it difficult to apply to limited hardware resources such as mobile or Internet of Things devices. To solve these limitations, network lightening studies have been actively conducted to reduce the depth or size of pre-trained deep CNN models while maintaining their performance as much as possible. This paper aims to lighten the SR CNN model, SRGAN, using the knowledge distillation among network lightening technologies; thus, it proposes four techniques with different methods of transferring the knowledge of the teacher network to the student network and presents experiments to compare and analyze the performance of each technique. In our experimental results, it was confirmed through quantitative and qualitative evaluation indicators that student networks with knowledge transfer performed better than those without knowledge transfer, and among the four knowledge transfer techniques, the technique of conducting adversarial learning after transferring knowledge from the teacher generator to the student generator showed the best performance.

Recently various attempts to control hardware through integration of sensors and artificial intelligence have been made. This paper proposes a smart hazardous sound detection at home. Previous sound recognition methods have problems due to the processing of background sounds and the low recognition accuracy of high-frequency sounds. To get around these problems, a new MFCC(Mel-Frequency Cepstral Coefficient) algorithm using Wiener filter, modified filterbank is proposed. Experiments for comparing the performance of the proposed method and the original MFCC were conducted. For the classification of feature vectors extracted using the proposed MFCC, DNN(Deep Neural Network) was used. Experimental results showed the superiority of the modified MFCC in comparison to the conventional MFCC in terms of 1% higher training accuracy and 6.6% higher recognition rate.

In this paper, the purpose is evaluation of the effect of using fractal feature in machine learning based pancreatic tumor classification. We used the data that Pancreas CT series 469 case including 1995 slice of benign and 1772 slice of malignant. Feature selection is implemented from 109 feature to 7 feature by Lasso regularization. In Fractal feature, fractal dimension is obtained by box-counting method, and hurst coefficient is calculated range data of pixel value in ROI. As a result, there were significant differences in both benign and malignancies tumor. Additionally, we compared the classification performance between model without fractal feature and model with fractal feature by using support vector machine. The train model with fractal feature showed statistically significant performance in comparison with train model without fractal feature.

With the continued development of image processing technology, we live in a time when it is difficult to visually discriminate processed (or tampered) images from real images. However, as the risk of fake images being misused for crime increases, the importance of image forensic science for identifying fake images is emerging. Currently, various deep learning-based identifiers have been studied, but there are still many problems to be used in real situations. Due to the inherent characteristics of deep learning that strongly relies on given training data, it is very vulnerable to evaluating data that has never been viewed. Therefore, we try to find a way to improve generalization ability of deep learning-based fake image identifiers. First, images with various contents were added to the training dataset to resolve the over-fitting problem that the identifier can only classify real and fake images with specific contents but fails for those with other contents. Next, color spaces other than RGB were exploited. That is, fake image identification was attempted on color spaces not considered when creating fake images, such as HSV and YCbCr. Finally, dropout, which is commonly used for generalization of neural networks, was used. Through experimental results, it has been confirmed that the color space conversion to HSV is the best solution and its combination with the approach of increasing the training dataset significantly can greatly improve the accuracy and generalization ability of deep learning-based identifiers in identifying fake images that have never been seen before.

In order to improve target tracking performance of the conventional electro-optical tracking system (EOTS) in the high speed and high maneuvering vehicle, an EOTS navigation algorithm is proposed, in which an inertial measurement unit(IMU) is included and navigation results of the vehicle are used. The proposed algorithm integrates vehicle's navigation results and the IMU and the time-delay and the scale factor errors are augmented into the integrated Kalman filter. In order to evaluate the proposed navigation algorithm, a land vehicle navigation experiments were performed a navigation grade navigation system, TALIN4000 and a tactical grade IMU, LN-200 and a equipment for roll motion were loaded on the land vehicle. The performance evaluation results show that the proposed algorithm effecting works in high maneuvering environment and for the time-delay.

To analyze the crime scene, the role of digital evidence such as CCTV and black box is very important. Such digital evidence is often damaged due to device defects or intentional deletion. In this case, the deleted video can be restored by well-known techniques like the frame-based recovery method. Especially, the data such as the video can be generally fragmented and saved in the case of the memory used almost fully. If the fragmented video were recovered in units of images, the sequence of the recovered images may not be continuous. In this paper, we proposed a new video restoration method to match the sequence of recovered images. First, the images are recovered through a frame-based recovery technique. Then, after analyzing the time information marked on the images, the time information was extracted and recognized via optical character recognition (OCR). Finally, the recovered images are rearranged based on the time information obtained by OCR. For performance evaluation, we evaluate the recovery rate of our proposed video restoration method. As a result, it was shown that the recovery rate for the fragmented video was recovered from a minimum of about 47% to a maximum of 98%.

In this paper, we propose a small waste appliance collection system that can be used for efficient waste appliance collection, and implement an unmanned automatic small waste appliance collector system. First, we collected cases related to the collection of waste home appliances, compared and analyzed methods, and arranged the characteristics. We analyzed at the requirements for the collection and management system of waste appliances by dividing it into the aspect of the discharger and the recycling business company. The collection system consists of a client-server structure, which is a collector system and a collection management server. The unmanned automatic collector system is divided into a control unit and a machine/mechanical unit. We identified the functions that the collector system must perform and the hardware required to perform these functions. Based on this, the collector system was implemented as an embedded system. The hardware and software used in the implementation are described and the implementation results are described. The collection system developed in this paper will contribute to the development of urban mining industry by enabling the efficient collection of waste appliances.

There are various typing practice applications. In addition, research cases on learning applications that support typing practice have been reported. These services are usually provided in a way that utilizes their own built-in text. Learners collect various contents through web services and use them a lot for learning. Therefore, this paper proposes a learning application to increase the learning effect by collecting vast amounts of web content and applying it to typing practice. The proposed application is implemented using Tkinter, a GUI module of Python. BeautifulSoup module of Python is used to extract information from the web. In order to process the extracted data, the NLTK module, which is an English data preprocessor, and the KoNLPy module, which is a Korean language processing module, are used. The operation of the proposed function is verified in the implementation and experimental results.

Item cold start is a well studied problem in the research field of recommender systems. Still, many existing collaborative filters cannot recommend items accurately when only a few user-item interaction data are available for newly introduced items (Cold items). We propose a interaction feature prediction method to mitigate item cold start problem. The proposed method predicts the interaction features that collaborative filters can calculate for the cold items. For prediction, in addition to content features of the cold-items used by state-of-the-art methods, our method exploits the interaction features of k-nearest content neighbors of the cold-items. An attention network is adopted to extract appropriate information from the interaction features of the neighbors by examining the contents feature similarity between the cold-item and its neighbors. Our evaluation on a real dataset CiteULike shows that the proposed method outperforms state-of-the-art methods 0.027 in Recall@20 metric and 0.023 in NDCG@20 metric.

Digital Twin, one of the core technologies of the Fourth Industrial Revolution, is attracting attention as a very suitable technology for disaster monitoring such as fires and earthquakes. In this paper, we implement a system equipped with UWB RTLS(Ultra-Wideband Real Time Location System), real-time remote control, and video streaming, which are element technologies for disaster monitoring digital twin. Since the proposed system structure is based on a cloud server, the actual location of the UWB indoor positioning-based client is transmitted to the user device in real time and stored on the cloud server for statistical and data analysis. In addition, we demonstrate through experiments that outliers occurs when the value of RSSI(Received Signal Strength Indicator) decreases due to communication collisions between UWB Tags, and propose an RSSI outlier correction algorithm to solve this problem.

The purpose of this study was to improve the experience of visiting obstetrics and gynecology of single women. After analyzing previous studies on Korean single women's perception of visiting obstetrics and gynecology, Contextual Interviews and Cultural Probes were conducted on single women in their 20s who visited obstetrics and gynecology. Based on this, personas were constructed to solidify the direction of problem solving by identifying the behavioral patterns and characteristics of single women. In this study, factors that hinder unmarried women's visits to obstetrics and gynecology and improvement measures were derived based on the information obtained using service design tools such as User Journey Mapping and Stakeholders' Map. Afterwards, a preference survey was conducted to increase the persuasiveness of the proposed method. The follow-up research task is to produce and propose the derived solution as a prototype that can be used in the actual field, and then proceed with user evaluation.