• IEIE Transactions on Smart Processing and Computing
• /
• v.1 no.3
• /
• pp.125-132
• /
• 2012

This paper proposes an automatic object segmentation and background composition method for video communication over consumer mobile phones. The object regions were extracted based on the motion and color variance of the first two frames. To combine the motion and variance information, the Euclidean distance between the motion boundary pixel and the neighboring color variance edge pixels was calculated, and the nearest edge pixel was labeled to the object boundary. The labeling results were refined using the morphology for a more accurate and natural-looking boundary. The grow-cut segmentation algorithm begins in the expanded label map, where the inner and outer boundary belongs to the foreground and background, respectively. The segmented object region and a new background image stored a priori in the mobile phone was then composed. In the background composition process, the background motion was measured using the optical-flow, and the final result was synthesized by accurately locating the object region according to the motion information. This study can be considered an extended, improved version of the existing background composition algorithm by considering motion information in a video. The proposed segmentation algorithm reduces the computational complexity significantly by choosing the minimum resolution at each segmentation step. The experimental results showed that the proposed algorithm can generate a fast, accurate and natural-looking background composition.

• Journal of Korea Multimedia Society
• /
• v.20 no.8
• /
• pp.1271-1281
• /
• 2017

It is essential to collect and analyze target information around the vehicle for autonomous driving of the vehicle. Based on the analysis, environmental information such as location and direction should be analyzed in real time to control the vehicle. In particular, obstruction or cutting of objects in the image must be handled to provide accurate information about the vehicle environment and to facilitate safe operation. In this paper, we propose a method to simultaneously generate 2D and 3D bounding box proposals using LiDAR Edge generated by filtering LiDAR sensor information. We classify the classes of each proposal by connecting them with Region-based Fully-Covolutional Networks (R-FCN), which is an object classifier based on Deep Learning, which uses two-dimensional images as inputs. Each 3D box is rearranged by using the class label and the subcategory information of each class to finally complete the 3D bounding box corresponding to the object. Because 3D bounding boxes are created in 3D space, object information such as space coordinates and object size can be obtained at once, and 2D bounding boxes associated with 3D boxes do not have problems such as occlusion.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.11 no.6
• /
• pp.2963-2978
• /
• 2017

The Top-K query is currently played a key role in a wide range of road network, decision making and quantitative financial research. In this paper, a Top-K recommendation algorithm is proposed to solve the cold-start problem and a tag generating method is put forward to enhance the semantic understanding of the OLAP session. In addition, a recommendation system for OLAP sessions called "OLAP4R" is designed using collaborative filtering technique aiming at guiding the user to find the ultimate goals by interactive queries. OLAP4R utilizes a mixed system architecture consisting of multiple functional modules, which have a high extension capability to support additional functions. This system structure allows the user to configure multi-dimensional hierarchies and desirable measures to analyze the specific requirement and gives recommendations with forthright responses. Experimental results show that our method has raised 20% recall of the recommendations comparing the traditional collaborative filtering and a visualization tag of the recommended sessions will be provided with modified changes for the user to understand.

• Proceedings of the Korean Vacuum Society Conference
• /
• 2013.02a
• /
• pp.85-85
• /
• 2013

Label-free, sensitive and selective detection methods with high spatial resolution are critically required for future applications in chemical sensor, biological sensor, and nanospectroscopic imaging. Here I describe the development of Plasmon Resonance Energy Transfer (PRET)-based molecular imaging in living cells as the first demonstration of intracellular imaging with PRET-based nanospectroscopy. In-vivo PRET imaging relied on the overlap between plasmon resonance frequency of gold nanoplasmonic probe (GNP) and absorption peak frequencies of conjugated molecules, which leads to create 'quantized quenching dips' in Rayleigh scattering spectrum of GNP. The position of these dips exactly matched with the absorption peaks of target molecules. As another innovative application of PRET, I present a highly selective and sensitive detection of metal ions by creating conjugated metal-ligand complexes on a single GNP. In addition to conferring high spatial resolution due to the small size of the metal ion probes (50 nm in diameter), this method is 100 to 1,000 folds more sensitive than organic reporter-based methods. Moreover, this technique achieves high selectivity due to the selective formation of Cu2+complexes and selective resonant quenching of GNP by the conjugated complexes. Since many metal ion ligand complexes generate new absorption peak due to the d-d transition in the metal ligand complex when a specific metal ion is inserted into the complex, we can match with the scattering frequency of nanoplasmonic metal ligand systems and the new absorption peak.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.12 no.3
• /
• pp.1264-1286
• /
• 2018

In this paper, a novel background prior-based salient object detection framework is proposed to deal with images those are more complicated. We take the superpixels located in four borders into consideration and exploit a mechanism based on image boundary information to remove the foreground noises, which are used to form the background prior. Afterward, an initial foreground prior is obtained by selecting superpixels that are the most dissimilar to the background prior. To determine the regions of foreground and background based on the prior of them, a threshold is needed in this process. According to a fixed threshold, the remaining superpixels are iteratively assigned based on their proximity to the foreground or background prior. As the threshold changes, different foreground priors generate multiple different partitions that are assigned a likelihood of being foreground. Last, all segments are combined into a saliency map based on the idea of similarity voting. Experiments on five benchmark databases demonstrate the proposed method performs well when it compares with the state-of-the-art methods in terms of accuracy and robustness.

• Journal of the Semiconductor & Display Technology
• /
• v.13 no.2
• /
• pp.45-52
• /
• 2014

This paper proposes an efficient photo retrieval system that automatically indexes for searching of relevant images, using a combination of geo-coded information, direction/location of image capture device and content-based visual features. A photo image is labeled with its GPS (Global Positioning System) coordinates and direction of the camera view at the moment of capture, and the label leads to generate a geo-spatial index with three core elements of latitude, longitude and viewing direction. Then, content-based visual features are extracted and combined with the geo-spatial information, for indexing and retrieving the photo images. For user's querying process, the proposed method adopts two steps as a progressive approach, filtering the relevant subset prior to use a content-based ranking function. To evaluate the performance of the proposed scheme, we assess the simulation performance in terms of average precision and F-score, using a natural photo collection. Comparing the proposed approach to retrieve using only visual features, an improvement of 20.8% was observed. The experimental results show that the proposed method exhibited a significant enhancement of around 7.2% in retrieval effectiveness, compared to previous work. These results reveal that a combination of context and content analysis is markedly more efficient and meaningful that using only visual feature for image search.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.13 no.4
• /
• pp.2223-2242
• /
• 2019

Distance Vector-Hop (DV-Hop) algorithm is widely used in node localization. It often suffers the wormhole attack. The current researches focus on Double-Wormhole-Node-Link (DWNL) and have limited attention to Multi-Wormhole-Node-Link (MWNL). In this paper, we propose a security DV-Hop algorithm (AMLDV-Hop) to resist MWNL. Firstly, the algorithm establishes the Neighbor List (NL) in initialization phase. It uses the NL to find the suspect beacon nodes and then find the actually attacked beacon nodes by calculating the distances to other beacon nodes. The attacked beacon nodes generate and broadcast the conflict sets to distinguish the different wormhole areas. The unknown nodes take the marked beacon nodes as references and mark themselves with different numbers in the first-round marking. If the unknown nodes fail to mark themselves, they will take the marked unknown nodes as references to mark themselves in the second-round marking. The unknown nodes that still fail to be marked are semi-isolated. The results indicate that the localization error of proposed AMLDV-Hop algorithm has 112.3%, 10.2%, 41.7%, 6.9% reduction compared to the attacked DV-Hop algorithm, the Label-based DV-Hop (LBDV-Hop), the Secure Neighbor Discovery Based DV-Hop (NDDV-Hop), and the Against Wormhole DV-Hop (AWDV-Hop) algorithm.

• Proceedings of the Korean Institute of Information and Commucation Sciences Conference
• /
• 2019.05a
• /
• pp.338-340
• /
• 2019

Recently, the speed and size of data accumulation are increasing due to the development of networks. There are many difficulties in classifying these data. One of the difficulties is the difficulty of labeling. Labeling is usually done by people, but it is very difficult for everyone to understand the data in the same way and it is very difficult to label them on the same basis. In order to solve this problem, we implemented GAN to generate new image based on input image and to learn input data indirectly by using it for learning. This suggests that the accuracy of classification can be increased by increasing the number of learning data.

• Journal of information and communication convergence engineering
• /
• v.21 no.1
• /
• pp.98-102
• /
• 2023

This paper presents a study on how augmenting semi-synthetic image data improves the performance of human detection algorithms. In the field of object detection, securing a high-quality data set plays the most important role in training deep learning algorithms. Recently, the acquisition of real image data has become time consuming and expensive; therefore, research using synthesized data has been conducted. Synthetic data haves the advantage of being able to generate a vast amount of data and accurately label it. However, the utility of synthetic data in human detection has not yet been demonstrated. Therefore, we use You Only Look Once (YOLO), the object detection algorithm most commonly used, to experimentally analyze the effect of synthetic data augmentation on human detection performance. As a result of training YOLO using the Penn-Fudan dataset, it was shown that the YOLO network model trained on a dataset augmented with synthetic data provided high-performance results in terms of the Precision-Recall Curve and F1-Confidence Curve.

• Proceedings of the Korea Water Resources Association Conference
• /
• 2023.05a
• /
• pp.26-26
• /
• 2023

General circulation models (GCMs) are widely used in hydrological prediction, however their coarse grids make them unsuitable for regional analysis, therefore a downscaling method is required to utilize them in hydrological assessment. As one of the downscaling methods, convolutional neural network (CNN)-based downscaling has been proposed in recent years. The aim of this study is to generate the process of dynamic downscaling using CNNs by applying GCM output as input and RCM output as label data output. Prediction accuracy is compared between different input datasets, and model structures. Several input datasets with key atmospheric variables such as precipitation, temperature, and humidity were tested with two different formats; one is two-dimensional data and the other one is three-dimensional data. And in the model structure, the hyperparameters were tested to check the effect on model accuracy. The results of the experiments on the input dataset showed that the accuracy was higher for the input dataset without precipitation than with precipitation. The results of the experiments on the model structure showed that substantially increasing the number of convolutions resulted in higher accuracy, however increasing the size of the receptive field did not necessarily lead to higher accuracy. Though further investigation is required for the application, this paper can contribute to the development of efficient downscaling method with CNNs.