In the characteristic of the brain malignant, the blood vessels and tumors have the same color and shape, and the boundary distinction is not clear, Therefore, it is difficult to observe the naked eye. Because of the high invasiveness, the risk of recurrence is high. Therefore, complete resection of the tumor is essential. The method for distinguishing the boundary between blood vessels and tumors is a fluorescence contrast method using indocyanine green (ICG), a fluorescence contrast agent. In ICG, the concentration range analysis is very important because the fluorescence expression state varies depending on the concentration. However, since the analysis result of the fluorescence expression condition is insufficient according to the current concentration, this paper proposes by analyzing the initial protocol of the concentration range. 780 nm infrared light was irradiated to the ICG sample to observe the fluorescence expression through a near infrared (NIR) camera. The wavelength is measured by using a spectrum instrument (ocean view) to observe the fluorescence expression wavelength of 811nm. As a result of analyzing the mol concentration according to each sample, the fluorescence expression range was found to be 16.15-0.16M and the optimum fluorescence concentration on the brightest part was found to be 3.23-0.81M.

Deep learning approach based on convolution neural network (CNN) has extensively studied in the field of computer vision. However, periocular feature extraction using CNN was not well studied because it is practically impossible to collect large volume of biometric data. This study uses the ResNet model which was trained with the ImageNet dataset. To overcome the problem of insufficient training data, we focused on the training of multi-layer perception (MLP) having simple structure rather than training the CNN having complex structure. It first extracts features using the pretrained ResNet model and reduces the feature dimension by principle component analysis (PCA), then trains a MLP classifier. Experimental results with the public periocular dataset UBIPr show that the proposed method is effective in person authentication using periocular region. Especially it has the advantage which can be directly applied for other biometric traits.

This paper proposes a novel active stereo matching technique using white auxiliary stripe pattern. The conventional active stereo matching techniques that uses two cameras and an active source such as projector can accurately estimate disparity information even in the areas with low texture compared to the passive ones. However, it is difficult that the conventional active stereo matching techniques using color code patterns acquire these patterns robustly if the object is composed of various colors or is exposed to complex lighting condition. To overcome this problem, the proposed method uses an additional white auxiliary stripe pattern to get and localize the color code patterns robustly. This paper proposes a process based on adaptive thresholding and thinning to obtain the auxiliary pattern accurately. Experimental results show that the proposed method more precisely measures the stepped sample whose depth is known in advance than the conventional method.

In this paper, an prediction method is proposed for coding of depth pictures using spherical modeling. An spherical surface which has the least error from original depth values is modeled in a block. Pixels in the block are predicted through the parameters of the modeled spherical surface. Simulation results show that average prediction errors and entropy powers are improved to 30% and 200% comparing to the intra prediction of H.264/AVC, selection ratios of the proposed spherical modeling mode is more than 25%.

The cosmetic and behavioral aspects of aging have become increasingly evident over the years. Physical aging in people can easily be observed on their face, posture, voice, and gait. In contrast, bone aging only becomes apparent once significant bone degeneration manifests through degenerative bone diseases. Therefore, a more accurate and timely assessment of bone aging is needed so that the determinants and its mechanisms can be more effectively identified and ultimately optimized. This study proposed a deep learning approach to assess the bone age of an adult using whole-body bone scintigraphy. The proposed approach uses multiple inputs deep neural network architectures using a loss function, called mean-variance loss. The data set was collected from Chonnam National University Hwasun Hospital. The experiment results show the effectiveness of the proposed method with a mean absolute error of 3.40 years.

In this paper, we propose a method to fabricate a patient-specific breast implant using MRI images and 3D scan data. Existing breast implants for breast reconstruction surgery are primarily fabricated products for shaping, and among the limited types of implants, products similar to the patient's breast have been used. In fact, the larger the difference between the shape of the breast and the implant, the more frequent the postoperative side effects and the lower the satisfaction. Previous researches on the fabrication of patient-specific breast implants have used limited information based on only MRI images or on only 3D scan data. In this paper, we propose an algorithm for the fabrication of patient-specific breast implants that combines MRI images with 3D scan data, considering anatomical suitability for external shape, volume, and pectoral muscle. Experimental results show that we can produce precise breast implants using the proposed algorithm.

The radar wind profiler data contaminates with various non-atmospheric components that produce errors in moments and wind velocity estimations. This study implemented an adaptive Gaussian model to detect and remove the clutter from the radar return. This model includes DC filtering, ground clutter recognition, Gaussian fitting, and cost function to mitigate the clutter component. The adaptive model tested for the various types of clutter components and found that it is effective in clutter removal process. It is also applied for the both time series and spectrum datasets. The moments estimated using this method are compared with those derived using conventional DC-filtering clutter removal method. The comparisons show that the proposed method effectively removes the clutter and produce reliable moments.

Educational books commonly use some copyrighted images with various kinds of deformation for helping students understanding. When using several copyrighted images made by merging or editing distortion in legal, we need to pay a charge to original copyright holders for each image. In this paper, we propose an efficient matching algorithm by separating each copyrighted image with the merged and edited type including rotation, illumination change, and change of size. We use the Oriented FAST and Rotated BRIEF (ORB) method as a basic feature matching scheme. To improve the matching accuracy, we design a new MIN-MAX similarity in matching stage. With the distorted dataset, the proposed method shows up-to 97% of precision in experiments. Also, we demonstrate that the proposed similarity measure also outperforms compared to other measure which is commonly used.

Nowadays object tracking process becoming one of the most challenging task in Computer Vision filed. A CSR-DCF (channel spatial reliability-discriminative correlation filter) tracking algorithm have been proposed on recent tracking benchmark that could achieve stat-of-the-art performance where channel spatial reliability concepts to DCF tracking and provide a novel learning algorithm for its efficient and seamless integration in the filter update and the tracking process with only two simple standard features, HoGs and Color names. However, there are some cases where this method cannot track properly, like overlapping, occlusions, motion blur, changing appearance, environmental variations and so on. To overcome that kind of complications a new modified version of CSR-DCF algorithm has been proposed by integrating deep learning based object detection and CSRT tracker which implemented in OpenCV library. As an object detection model, according to the comparable result of object detection methods and by reason of high efficiency and celerity of Faster RCNN (Region-based Convolutional Neural Network) has been used, and combined with CSRT tracker, which demonstrated outstanding real-time detection and tracking performance. The results indicate that the trained object detection model integration with tracking algorithm gives better outcomes rather than using tracking algorithm or filter itself.

In this paper, we proposed a methodology for estimating rainfall intensity using a W-band FMCW automotive radar signal which is the core technology of autonomous driving car. By comparing and analyzing the results of rainfall and non-rainfall observation, we found that the reflection intensity of the automotive radar is changed with rainfall intensity. We could confirm the possibility of deriving the quantitative precipitation estimation using the methodology derived from this result. In addition it can be possible to develop a new paradigm of precipitation observation technique by observing various events together with the weather radar and the ground rainfall observation equipment.

Recently, the number of mobile users as well as high-speed Internet user has been increasing rapidly. Moreover, traffic is growing fast as services that provide real-time content such as Youtube and Netflix become popular. The problem of traffic control in real-time content services is important because many people use cell phones to receive real-time content. In this regard, the field of CCN is currently being studied. We studied the mobility of nodes among CCN research fields. Node mobility can be divided into three categories : consumer mobility, intermediate node, and provide mobility. In this paper, we propose Mobility Node Support(MD-INS) to support the intermediate-node mobility in CCN real-time services. Experimental results show that the proposed scheme shows better performance than CCN in terms of service disconnection time and packet loss.

As artificial intelligence technology is actively used in image processing, it is possible to generate high-quality fake images based on deep learning. Fake images generated using GAN(Generative Adversarial Network), one of unsupervised learning algorithms, have reached levels that are hard to discriminate from the naked eye. Detecting these fake images is required as they can be abused for crimes such as illegal content production, identity fraud and defamation. In this paper, we develop a deep-learning model based on CNN(Convolutional Neural Network) for the detection of StyleGAN fake images. StyleGAN is one of GAN algorithms and has an excellent performance in generating face images. We experiment with 48 number of experimental scenarios developed by combining parameters of the proposed model. We train and test each scenario with 300,000 number of real and fake face images in order to present a model parameter that improves performance in the detection of fake faces.

3D convolution is to stack multiple consecutive frames to form a cube, and then apply the 3D convolution kernel in the cube. In this structure, each feature map of the convolutional layer is connected to multiple adjacent sequential frames in the previous layer, thus capturing the motion information. However, due to the changes of pedestrian posture, motion and position, the convolution at the same place is inappropriate, and when the 3D convolution kernel is convoluted in the time domain, only time domain features of three consecutive frames can be extracted, which is not a good enough to get action information. This paper proposes an action recognition method based on feature fusion of 3D convolutional neural network. Based on the VGG16 network model, sending a pre-acquired optical flow image for learning, then get the time domain features, and then the feature of the time domain is extracted from the features extracted by the 3D convolutional neural network. Finally, the behavior classification is done by the SVM classifier.

This paper addresses the efficient processing of all-farthest-neighbors (AFN) queries in spatial network databases. Given a set of data points P={p₁,p₂,…,p_|p|} in a spatial network, where the distance between two data points p and s, denoted by dist (p,s), is the length of the shortest path between them, an AFN query is defined as follows: find the farthest neighbor ω(p)∈P of each data point p such that dist(p,ω(p)) ≥ dist(p,s) for all s∈P. In this paper, we propose a shared execution algorithm called FAST (for All-Farthest-neighbors Search in spatial neTworks). Extensive experiments on real-world roadmaps confirm the efficiency and scalability of the FAST algorithm, while demonstrating a speedup of up to two orders of magnitude over a conventional solution.

Arduino makes it easy to connect objects and computers. As a result, programming learning using physical computing has been proposed as an effective alternative to SW training for beginners. In this paper, we propose an Arduino-based physical computing education method that can be applied to basic programming subjects. To this end, we propose a basic programming training method based on Arduino analog signals. Currently, physical computing courses focus on digital control when connecting input sensors and output devices in Arduino. However, the contents of programming education using analog signals of Arduino boards are insufficient. In this paper, we proposed and tested the teaching method used for programming education using low-cost materials used for Arduino analog signal-based computing.

Insufficient sleep time and bad sleep quality causes many illnesses and it's research became more and more important. The most common method for measuring sleep quality is the polysomnography(PSG). The PSG is a test used to diagnose sleep disorders. The most common PSG data is obtained from the examiner, which attaches several sensors on a body and takes sleep overnight. However, most of the sleep stage classification in PSG are low accuracy of the classification. In this paper, we have studied algorithm for sleep level classification based on machine learning which can replace PSG. EEG, EOG, and EMG channel signals are studied and tested by using CNN algorithm. In order to compensate the performance, a mixed model using both CNN and DNN models is designed and tested for performance.