Most of the previous studies for bark recognition have focused on the extraction of LBP-like statistical features. Deep learning approach was not well studied because of the difficulty of acquiring large volume of bark image dataset. To overcome the bark dataset problem, this study utilizes the MobileNet which was trained with the ImageNet dataset. This study proposes two approaches. One is to extract features by the pixel-wise convolution and classify the features with SVM. The other is to tune the weights of the MobileNet by flexibly freezing layers. The experimental results with two public bark datasets, BarkTex and Trunk12, show that the proposed methods are effective in bark recognition. Especially the results of the flexible tunning method outperform state-of-the-art methods. In addition, it can be applied to mobile devices because the MobileNet is compact compared to other deep learning models.

Reconstruction of the three-dimensional (3D) environment is a key aspect of augmented reality and augmented virtuality, which utilize and incorporate a user's surroundings. Such reconstruction can be easily realized by employing a Kinect device. However, multiple Kinect devices are required for enhancing the reconstruction density and for spatial expansion. While employing multiple Kinect devices, they must be calibrated with respect to each other in advance, and a marker is often used for this purpose. However, a marker needs to be placed at each calibration, and the result of marker detection significantly affects the calibration accuracy. Therefore, a user-friendly, efficient, accurate, and marker-less method for calibrating multiple Kinect devices is proposed in this study. The proposed method includes a joint tracking algorithm for approximate calibration, and the obtained result is further refined by applying the iterative closest point algorithm. Experimental results indicate that the proposed method is a convenient alternative to conventional marker-based methods for calibrating multiple Kinect devices. Hence, the proposed method can be incorporated in various applications of augmented reality and augmented virtuality that require 3D environment reconstruction by employing multiple Kinect devices.

We present a novel deep learning architecture for obtaining a latent image from a single blurry image, which contains dynamic motion blurs through object/camera movements. The proposed architecture consists of two sub-modules: blur image restoration and optical flow estimation. The tasks are highly related in that object/camera movements make cause blurry artifacts, whereas they are estimated through optical flow. The ablation study demonstrates that training multi-task architecture simultaneously improves both tasks compared to handling them separately. Objective and subjective evaluations show that our method outperforms the state-of-the-arts deep learning based techniques.

Because of the dynamic range limitation of digital equipment, it is impossible to obtain dark and bright areas at the same time with one shot. In order to solve this problem, an exposure fusion technique for fusing a plurality of images photographed at different exposure amounts into one is being studied. Among them, Laplacian pyramid decomposition based fusion method can generate natural HDR image by fusing images of various scales. But this requires a lot of computation time. Therefore, in this paper, we propose an approximation technique that achieves similar performance and greatly shortens computation time. The concept of vanishing point image for approximation is introduced, and the validity of the proposed approach is verified by comparing the computation time with the resultant image.

In order to diagnose and prevent Alzheimer's Disease (AD), it is becoming increasingly important to develop a CAD(Computer-aided Diagnosis) system for AD diagnosis, which provides effective treatment for patients by analyzing 3D MRI images. It is essential to apply powerful deep learning algorithms in order to automatically classify stages of Alzheimer's Disease and to develop a Alzheimer's Disease support diagnosis system that has the function of detecting hippocampus and CSF(Cerebrospinal fluid) which are important biomarkers in diagnosis of Alzheimer's Disease. In this paper, for AD diagnosis, we classify a given MRI data into three categories of AD, mild cognitive impairment, and normal control according by applying 3D brain MRI image to the Faster R-CNN model and detect hippocampus and CSF in MRI image. To do this, we use the 2D MRI slice images extracted from the 3D MRI data of the Faster R-CNN, and perform the widely used majority voting algorithm on the resulting bounding box labels for classification. To verify the proposed method, we used the public ADNI data set, which is the standard brain MRI database. Experimental results show that the proposed method achieves impressive classification performance compared with other state-of-the-art methods.

In recent years, interest in drones has grown, and many countries are developing them into a strategic industry of the future. Drones are not only used in industries such as logistics and agriculture but also in various public sectors such as life rescue, disaster investigation, traffic control, and firefighting. One of the most important tasks of a drone is to accurately identify targets in these applications. Target recognition may vary depending on the search environment of the drone. Therefore, this study tests and analyzes the drone's target recognition performance according to changes in the search environment such as the search altitude and the search angle. In addition, we propose a new algorithm that improves upon the disadvantages of the Haar cascade method, which is the existing algorithm that recognizes the target by analyzing a captured image.

In this study, we propose a new $PM_{10}$ forecasting model for Seoul region using DNN(Deep Neural Network) and secondary data. The previous numerical and Julian forecast model have been developed using primary data such as weather and air quality measurements. These models give excellent results for accuracy and false alarms, but POD is not good for the daily life usage. To solve this problem, we develop four secondary factors composed with primary data, which reflect the correlations between primary factors and high $PM_{10}$ concentrations. The proposed 4 models are A(Anomaly), BT(Back trajectory), CB(Contribution), CS(Cosine similarity), and ALL(model using all 4 secondary data). Among them, model ALL shows the best performance in all indicators, especially the PODs are improved.

Video surveillance data, which is used for preemptive or post-emptive action against any event or accident, is required for monitoring the location, but is reducing the capacity of the image data by removing intervals for cost reduction and system persistence. Such a video surveillance system is fixed in a certain position and monitors the area only within a limited angle, or monitors only the fixed area without changing the angle. At this time, the video surveillance system that is monitored only within a limited angle shows that the variation object such as the floating population shows different status in the image, and the background of the image maintains a generally constant appearance. The static objects in the image do not need to be stored in all the images, unlike the dynamic objects that must be continuously shot, and occupy a storage space other than the necessary ones. In this paper, we propose a mechanism to analyze the image, store only the small size image for the fixed background, and store it as image data only for variable objects.

4th industrial age! 5G-based video content has changed rapidly. How do you view and understand VR/AR/MR/holograms in the age of convergent technology media represented by digital convergence in the 21st century beyond the modern visual system of Descartes? It predicts extension to an interactive visual system. This study proposes a methodology to visualize the Lacanian visual acuity after applying it to the L-System and to understand the visual system of the 21st century.