• Journal of Broadcast Engineering
• /
• v.23 no.3
• /
• pp.403-408
• /
• 2018

Generally, a large data set is required for learning of deep learning. However, since it is not easy to create large data sets, there are a lot of techniques that make small data sets larger through data expansion such as rotation, flipping, and filtering. However, these simple techniques have limitation on extendibility because they are difficult to escape from the features already possessed. In order to solve this problem, we propose a method to acquire new image data by using existing data. This is done by retrieving and acquiring similar images using existing image data as a query of the content-based image retrieval (CBIR). Finally, we compare the performance of the base model with the model using CBIR.

• Communications for Statistical Applications and Methods
• /
• v.29 no.2
• /
• pp.161-176
• /
• 2022

There has been active research in image classification using deep learning convolutional neural network (CNN) models. ImageNet large-scale visual recognition challenge (ILSVRC) (2010-2017) was one of the most important competitions that boosted the development of efficient deep learning algorithms. This paper introduces and compares six monumental models that achieved high prediction accuracy in ILSVRC. First, we provide a review of the models to illustrate their unique structure and characteristics of the models. We then compare those models under a unified framework. For this reason, additional devices that are not crucial to the structure are excluded. Four popular data sets with different characteristics are then considered to measure the prediction accuracy. By investigating the characteristics of the data sets and the models being compared, we provide some insight into the architectural features of the models.

• Korean Journal of Remote Sensing
• /
• v.33 no.4
• /
• pp.339-350
• /
• 2017

In recent, cloud computing paradigm and open source as a huge trend in the Information Communication Technology (ICT) are widely applied, being closely interrelated to each other in the various applications. The integrated services by both technologies is generally regarded as one of a prospective web-based business models impacting the concerned industries. In spite of progressing those technologies, there are a few application cases in the geo-based application domains. The purpose of this study is to develop a cloud-based service system for satellite image processing based on the pure and full open source. On the OpenStack, cloud computing open source, virtual servers for system management by open source stack and image processing functionalities provided by OTB have been built or constructed. In this stage, practical image processing functions for KOMPSAT within this service system are thresholding segmentation, pan-sharpening with multi-resolution image sets, change detection with paired image sets. This is the first case in which a government-supporting space science institution provides cloud-based services for satellite image processing functionalities based on pure open source stack. It is expected that this implemented system can expand with further image processing algorithms using public and open data sets.

• Journal of Institute of Control, Robotics and Systems
• /
• v.16 no.7
• /
• pp.695-702
• /
• 2010

This paper presents an image alignment algorithm for application of AOI (Automatic Optical Inspection) based on SIFT. Since the correspondences result using SIFT descriptor have many wrong points for aligning, this paper modified and classified those points by five measures called the CCFMR (Cascade Classifier for False Matching Reduction) After reduced the false matching, rotation and translation are estimated by point selection method. Experimental results show that the proposed method has fewer fail matching in comparison to commercial software MIL 8.0, and specially, less than twice with the well-controlled environment’s data sets (such as AOI system). The rotation and translation accuracy is robust than MIL in the noise data sets, but the errors are higher than in a rotation variation data sets although that also meaningful result in the practical system. In addition to, the computational time consumed by the proposed method is four times shorter than that by MIL which increases linearly according to noise.

• Journal of the Korean Society of Clothing and Textiles
• /
• v.44 no.2
• /
• pp.354-368
• /
• 2020

Accurate and versatile image data-sets are essential for fashion AI research and AI-based fashion businesses based on a systematic attribute classification system. This study constructs a color and texture attribute hierarchical classification system by collecting fashion item images and analyzing the metadata of fashion items described by consumers. Essential dimensions to explain color and texture attributes were extracted; in addition, attribute values for each dimension were constructed based on metadata and previous studies. This hierarchical classification system satisfies consistency, exclusiveness, inclusiveness, and flexibility. The image tagging to confirm the usefulness of the proposed classification system indicated that the contents of attributes of the same image differ depending on the annotator that require a clear standard for distinguishing differences between the properties. This classification system will improve the reliability of the training data for machine learning, by providing standardized criteria for tasks such as tagging and annotating of fashion items.

• Korean Journal of Remote Sensing
• /
• v.38 no.1
• /
• pp.21-32
• /
• 2022

The importance of the classical theme of how the Top-of-Atmosphere (TOA) and Top-of-Canopy (TOC) reflectance of high-resolution satellite images match the actual atmospheric reflectance and surface reflectance has been emphasized. Based on the Radiometric Calibration Network (RadCalNet) BTCN and BSCN data, this study compared the accuracy of TOA and TOC reflectance products of the currently available optical satellites, including KOMPSAT-3, WorldView-2, and Pléiades-1A image sets calculated using the absolute atmospheric correction function of the Orfeo Toolbox (OTB) tool. The comparison experiment used data in 2018 and 2019, and the Landsat-8 image sets from the same period were applied together. The experiment results showed that the product of TOA and TOC reflectance obtained from the three sets of images were highly consistent with RadCalNet data. It implies that any imagery may be applied when high-resolution reflectance products are required for a certain application. Meanwhile, the processed results of the OTB tool and those by the Apparent Reflection method of another tool for WorldView-2 images were nearly identical. However, in some cases, the reflectance products of Landsat-8 images provided by USGS sometimes showed relatively low consistency than those computed by the OTB tool, with the reference of RadCalNet BTCN and BSCN data. Continuous experiments on active vegetation areas in addition to the RadCalNet sites are necessary to obtain generalized results.

• Journal of Navigation and Port Research
• /
• v.27 no.3
• /
• pp.273-282
• /
• 2003

This paper proposes a sensor-fusion technique where rho data sets for the previous moments are properly transformed and fused into the current data sets to enable accurate measurement, such as, distance to an obstacle and location of the service robot itself. In the conventional fusion schemes, the measurement is dependent only on the current data sets. As the results, more of sensors are required to measure a certain physical promoter or to improve the accuracy of the measurement. However, in this approach, intend of adding more sensors to the system, the temporal sequence of the data sets are stored and utilized for the measurement improvement. Theoretical basis is illustrated by examples md the effectiveness is proved through the simulation. Finally, the new space and time sensor fusion (STSF) scheme is applied to the control of a mobile robot in the indoor environment and the performance was demonstrated by the real experiments.

• Journal of Korea Multimedia Society
• /
• v.22 no.9
• /
• pp.1020-1028
• /
• 2019

Cerebrovascular disease is a common disease in the elderly population. However, we do not have enough understanding of brain-related diseases. Recent advances in microscopy technology have resulted in the acquisition of vast amounts of image data sets for small organs, and it has become possible to handle vast amounts of image data sets due to improved computer performance and software technology. In this paper, the author proposes introduce a method for classifying and analysing only cerebrovascular information in the mouse brain image, as well as a quantitative measure of the portion of the cerebrovascular in the mouse brain. The study of the cerebrovascular structure is significant, and it can be helpful to improve the understanding of cerebrovasculature. As a result, the author expects that this study will be useful for neuroscientists conducting clinical research.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.9 no.2
• /
• pp.775-792
• /
• 2015

Person re-identification is an important and challenging task in computer vision with numerous real world applications. Despite significant progress has been made in the past few years, person re-identification remains an unsolved problem. This paper presents a novel appearance-based approach to person re-identification. The approach exploits region covariance matrix and color histograms to capture the statistical properties and chromatic information of each object. Robustness against low resolution, viewpoint changes and pose variations is achieved by a novel signature, that is, the combination of Log Covariance Matrix feature and HSV histogram (LCMH). In order to further improve re-identification performance, third-party image sets are utilized as a common reference to sufficiently represent any image set with the same type. Distinctive and reliable features for a given image set are extracted through decision boundary between the specific set and a third-party image set supervised by max-margin criteria. This method enables the usage of an existing dataset to represent new image data without time-consuming data collection and annotation. Comparisons with state-of-the-art methods carried out on benchmark datasets demonstrate promising performance of our method.

• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.25 no.12
• /
• pp.1790-1796
• /
• 2021

A large amount of image data sets are required for image deep learning, and there are many differences in the method of obtaining images and constructing image data sets depending on the type of object. In this paper, we presented a method of constructing an image data set for deep learning and analyzed the performance that varies depending on the object to be tracked. We took a video by rotating the object, and then created a data set by segmenting the video using the proposed data set construction method. As a result of performance analysis, detection rate was more than 95%, and detection rate of objects with little change in shape was higher performance. It is considered that it is effective to use the data set construction method presented in this paper for a situation in which it is difficult to obtain image data and to track an object with little change in shape within a video.