• The KIPS Transactions:PartB
• /
• v.14B no.2
• /
• pp.65-70
• /
• 2007

Red tide is becoming hot issue of environmental problem worldwide since the 1990. Advanced nations, are progressing study that detect red tide area on early time using satellite for sea. But, our country most seashores bends serious. Also because there are a lot of turbid method streams on coast, hard to detect small red tide area by satellite for sea that is low resolution. Also, method by sea color that use one feature of satellite image for sea of existent red tide area detection was most. In this way, have a few feature in image with sea color and it can cause false negative mistake that detect red tide area. Therefore, in this paper, acquired texture information to use GLCM(Gray Level Co occurrence Matrix)'s texture 6 information about high definition land satellite south Coast image. Removed needless component reducing dimension through principal component analysis from this information. And changed into 2 principal component accumulation images, Experiment result 2 principal component conversion accumulation image's eigenvalues were 94.6%. When component with red tide area that uses only sea color image and all principal component image. displayed more correct result. And divided as quantitative,, it compares with turbid stream and the sea that red tide does not exist using statistical feature analysis about texture.

• KIPS Transactions on Software and Data Engineering
• /
• v.6 no.4
• /
• pp.203-210
• /
• 2017

In this paper, we propose an effective neural network model for image caption generation and model transfer. This model is a kind of multi-modal recurrent neural network models. It consists of five distinct layers: a convolution neural network layer for extracting visual information from images, an embedding layer for converting each word into a low dimensional feature, a recurrent neural network layer for learning caption sentence structure, and a multi-modal layer for combining visual and language information. In this model, the recurrent neural network layer is constructed by LSTM units, which are well known to be effective for learning and transferring sequence patterns. Moreover, this model has a unique structure in which the output of the convolution neural network layer is linked not only to the input of the initial state of the recurrent neural network layer but also to the input of the multimodal layer, in order to make use of visual information extracted from the image at each recurrent step for generating the corresponding textual caption. Through various comparative experiments using open data sets such as Flickr8k, Flickr30k, and MSCOCO, we demonstrated the proposed multimodal recurrent neural network model has high performance in terms of caption accuracy and model transfer effect.

• Journal of the Korean Institute of Intelligent Systems
• /
• v.26 no.1
• /
• pp.56-63
• /
• 2016

In this study, we introduce a design of Fuzzy RBFNNs-based digit recognition system using the incremental-PCA in order to recognize the handwritten digits. The Principal Component Analysis (PCA) is a widely-adopted dimensional reduction algorithm, but it needs high computing overhead for feature extraction in case of using high dimensional images or a large amount of training data. To alleviate such problem, the incremental-PCA is proposed for the computationally efficient processing as well as the incremental learning of high dimensional data in the feature extraction stage. The architecture of Fuzzy Radial Basis Function Neural Networks (RBFNN) consists of three functional modules such as condition, conclusion, and inference part. In the condition part, the input space is partitioned with the use of fuzzy clustering realized by means of the Fuzzy C-Means (FCM) algorithm. Also, it is used instead of gaussian function to consider the characteristic of input data. In the conclusion part, connection weights are used as the extended diverse types in polynomial expression such as constant, linear, quadratic and modified quadratic. Experimental results conducted on the benchmarking MNIST handwritten digit database demonstrate the effectiveness and efficiency of the proposed digit recognition system when compared with other studies.

• Cartoon and Animation Studies
• /
• s.45
• /
• pp.27-50
• /
• 2016

The present study argues that documentary-animation films, which are based on actual human voices, on the level of representation, constitute a new expansion for the medium of animation films, which serve as testimonies to the real world. Animation films are produced using very diverse techniques so that they are complex to the degree of being indefinable, and documentary films, though based on objective representation, increase in complexity in that there exist various types of artificial interventions such as direction and digital image processing. Having emerged as a hybrid genre of the two media, documentary-animation films draw into themselves actual events and elements so that they conceptually share reality-based narratives and are visually characterized by the trappings of animation films. Generally classified as 'animated documentaries', this genre triggered discussions following the release of , a work that is mistaken as having used rotoscoping transforming live action in terms of the technique. When analyzed in detail, however, this work is presented as an ambiguous medium where the characteristics of animation films, which are virtual simulacra without reality, and of documentaries, which are based on the objective indexicality of the referents, coexist because of its mixed use of typical animation techniques, 3D programs, and live-action images. Discussed in the present study, , , and share the characteristics of the medium of documentaries in that the narratives develop as testimonies of historical figures but, at the same time, are connected to animation films because of their production techniques and direction characteristics. Consequently, this medium must be discussed as a new expansion rather than being included in the existing classification system, and such a presupposition is an indispensable process for directly facing the reality of the works and for developing discussions. Through works that directly use the interviewees' voices yet do not transcend the characteristics of animation films, the present study seeks to define documentary-animation films and to discuss the possibility of the medium, which has expanded as a testimony to the real world.

• Tunnel and Underground Space
• /
• v.26 no.1
• /
• pp.24-31
• /
• 2016

Recently, many studies have been conducted to use fixed-wing and rotary-wing unmanned aerial vehicles (UAVs, Drones) for topographic surveying in open-pit mines. Because the fixed-wing and rotary-wing UAVs have different characteristics such as flight height, speed, time and performance of mounted cameras, their results of topographic surveying at a same site need to be compared. This study selected a construction site in Yangsan-si, Gyeongsangnam-do, Korea as a study area and compared the topographic surveying results from a fixed-wing UAV (SenseFly eBee) and a popular rotary-wing UAV (DJI Phantom2 Vision+). As results of data processing for aerial photos taken from eBee and Phantom2 Vision+, orthomosaic images and digital surface models with about 4 cm grid spacing could be generated. Comparisons of the X, Y, Z-coordinates of 7 ground control points measured by differential global positioning system and those determined by eBee and Phantom2 Vision+ revealed that the root mean squared errors of X, Y, Z-coordinates were around 10 cm, respectively.

• Korean Journal of Remote Sensing
• /
• v.36 no.5_3
• /
• pp.1095-1107
• /
• 2020

Flood monitoring using satellite data has been constrained by obtaining satellite images for flood peak and accurately extracting flooded areas from satellite data. Deep learning is a promising method for satellite image classification, yet the potential of deep learning-based flooded area extraction using SAR data remained uncertain, which has advantages in obtaining data, comparing to optical satellite data. This research explores the performance of SegNet and U-Net on image segmentation by extracting flooded areas in the Khorat basin, Mekong river basin, and Cagayan river basin in Thailand, Laos, and the Philippines from Sentinel-1 A/B satellite data. Results show that Global Accuracy, Mean IoU, and Mean BF Score of SegNet are 0.9847, 0.6016, and 0.6467 respectively, whereas those of U-Net are 0.9937, 0.7022, 0.7125. Visual interpretation shows that the classification accuracy of U-Net is higher than SegNet, but overall processing time of SegNet is around three times faster than that of U-Net. It is anticipated that the results of this research could be used when developing deep learning-based flood monitoring models and presenting fully automated flooded area extraction models.

• Korean Journal of Remote Sensing
• /
• v.34 no.2_2
• /
• pp.407-418
• /
• 2018

The frequency and risk of disasters are increasing due to environmental and social factors. In order to respond effectively to disasters that occur unexpectedly, it is very important to quickly obtain up-to-date information about target area. It is possible to intuitively judge the situation about the area through the image-map generated at high speed, so that it can cope with disaster quickly and effectively. In this study, we propose an image-map generation and visualization system from UAV images for real-time disaster monitoring. The proposed system consists of aerial segment and ground segment. In the aerial segment, the UAV system acquires the sensory data from digital camera and GPS/IMU sensor. Communication module transmits it to the ground server in real time. In the ground segment, the transmitted sensor data are processed to generate image-maps and the image-maps are visualized on the geo-portal. We conducted experiment to check the accuracy of the image-map using the system. Check points were obtained through ground survey in the data acquisition area. When calculating the difference between adjacent image maps, the relative accuracy was 1.58 m. We confirmed the absolute accuracy of the image map for the position measured from the individual image map. It is confirmed that the map is matched to the existing map with an absolute accuracy of 0.75 m. We confirmed the processing time of each step until the visualization of the image-map. When the image-map was generated with GSD 10 cm, it took 1.67 seconds to visualize. It is expected that the proposed system can be applied to real - time monitoring for disaster response.

• Journal of the Korean Society of Radiology
• /
• v.1 no.1
• /
• pp.25-30
• /
• 2007

ROC(Receiver Operating Characteristic)curve is the method that estimate detected insignificant signal from the human's sense of sight, it has been raised excellent results. In this study, we evaluate image quality and equipment character by obtaining a chest image from CR(Computed Radiography) and DDR(Direct Digital radiography) using the human chest phantom, The parameter of exposure for obtaining chest image was 120 kVp/3.2 mAs and the SID(Source to Image Distance) was 180cm. The images were obtained by CR(AGFA MD 4.0 General plate, JAPAN) and DDR(HOLOGIC nDirect Ray, USA). Using some pieces of Aluminum and stone for expressing regions, then attached them on the heart, lung and thoracic vertebrae of the phantom. 29 persons hold radiology degrees were participated in ROC analysis. As a result of the ROC analysis, TPF(true positive fraction) and FPF(false positive fraction) of DDR and CR are 0.552 and 0.474 and 0.629 and 0.405, respectively. By using the results, the ROC curve of CR has higher image quality than DDR. According to the theory, DDR has the higher image quality than CR in chest X-ray image. But, CR has the higher image quality than DDR. quality of DDR inserted the enhance board. The results confirmed that image post-processing is important element decipherment of clinical.

• Journal of KIISE:Computer Systems and Theory
• /
• v.32 no.10
• /
• pp.530-540
• /
• 2005

Virtual colonoscopy is an easy and fast method to reconstruct the shape of colon and diagnose tumors inside the colon based on computed tomography images. This is a non-invasive method, which resolves weak points of previous invasive methods. The path for virtual colonoscopy should be generated rapidly and accurately for clinical examination. However, previous methods are computationally expensive because the data structure such as distance map should be constructed in the preprocessing and positions of all the points of the path needs to be calculated. In this paper, we propose the automatic path generation method based on visibility to decrease path generation time. The proposed method does not require preprocessing and generates small number of control points representing the Path instead of all points to generate the path rapidly. Also, our method generates the path based on visibility so that a virtual camera moves smoothly and a comfortable and accurate path is calculated for virtual navigation. Also, our method can be used for general virtual navigation of various kinds of pipes.

• Journal of KIISE:Computing Practices and Letters
• /
• v.9 no.6
• /
• pp.646-661
• /
• 2003

It is not easy that we visualize the large volume data stored in the every client computers of the web environment. One solution is as follows. First we compress volume data, second store that in the database server, third transfer that to client computer, fourth visualize that with direct-volume-rendering in the client computer. In this case, we usually use wavelet transform for compressing large data. This paper reports the experiments for acquiring the wavelet bases and the compression ratios fit for the above processing paradigm. In this experiments, we compress the volume data Engine, CThead, Bentum into 50%, 10%, 5%, 1%, 0.1%, 0.03% of the total data respectively using Harr, Daubechies4, Daubechies12 and Daubechies20 wavelets, then visualize that with direct-volume-rendering, afterwards evaluate the images with eyes and image comparison metrics. When compression ratio being low the performance of Harr wavelet is better than the performance of the other wavelets, when compression ratio being high the performance of Daubechies4 and Daubechies12 is better than the performance of the other wavelets. When measuring with eyes the good compression ratio is about 1% of all the data, when measuring with image comparison metrics, the good compression ratio is about 5-10% of all the data.