• Smart Structures and Systems
• /
• v.29 no.1
• /
• pp.195-206
• /
• 2022

Crack detection is essential for inspection of existing structures and crack segmentation based on deep learning is a significant solution. However, datasets are usually one of the key issues. When building a new dataset for deep learning, laborious and time-consuming annotation of a large number of crack images is an obstacle. The aim of this study is to develop an approach that can automatically select a small portion of the most informative crack images from a large pool in order to annotate them, not to label all crack images. An active learning method with difficulty learning mechanism for crack segmentation tasks is proposed. Experiments are carried out on a crack image dataset of a steel box girder, which contains 500 images of 320×320 size for training, 100 for validation, and 190 for testing. In active learning experiments, the 500 images for training are acted as unlabeled image. The acquisition function in our method is compared with traditional acquisition functions, i.e., Query-By-Committee (QBC), Entropy, and Core-set. Further, comparisons are made on four common segmentation networks: U-Net, DeepLabV3, Feature Pyramid Network (FPN), and PSPNet. The results show that when training occurs with 200 (40%) of the most informative crack images that are selected by our method, the four segmentation networks can achieve 92%-95% of the obtained performance when training takes place with 500 (100%) crack images. The acquisition function in our method shows more accurate measurements of informativeness for unlabeled crack images compared to the four traditional acquisition functions at most active learning stages. Our method can select the most informative images for annotation from many unlabeled crack images automatically and accurately. Additionally, the dataset built after selecting 40% of all crack images can support crack segmentation networks that perform more than 92% when all the images are used.

• Journal of Korea Multimedia Society
• /
• v.20 no.12
• /
• pp.1992-1999
• /
• 2017

This study proposes a decorative typeface which is produced through the concept of trigonometric functions in an open-source programming language known as Processing. First, the theoretical background of Processing and trigonometric functions as well as previous research in this area are analyzed. Second, basic modules of 'V', 'I', 'O', and 'M' were created for use as the final alphabet typeface with the concept of a trigonometric function. Third, a decorative parabolic curve that encircles the base module was created. Finally, the modules created on Processing were edited in Adobe Illustrator to create a typeface set with characters from A to Z. Various artworks using Programming can produce an infinite number of different versions by modifying only some of the variables and codes, and this method can include multimedia features such as text, images, videos, interactive art and various forms of content and media. Therefore, with regard to expression, the possibilities are endless. In this study, I attempt to expand the field of visual culture using programming and computational methodologies. In contrast to the digital typeface production method, which relies on existing graphic tools, this study is meaningful because it expands the range of use of decorative typefaces.

• Proceedings of the Korea Contents Association Conference
• /
• 2015.05a
• /
• pp.191-192
• /
• 2015

전립선비대증(Benign Prostatic Hyperplasia, BPH)은 전립선조직중에 이행구역의 결절성증식과 요도 주위의 과증식(Hyperplasia)이 특징이다. 경직장초음파(TRUS: transrectal ultrasonography)검사를 이용한 진단에 있어 정상조직과 비대되어 있는 조직의 영상 차이를 비교하고 수량화로 나타내었다, 영상분석에는 GLCM 통계적 파라미터 중에서 Autocorrelation, Cluster Prominence, Entropy, Sum average를 4개의 파라미터에서 병변 인식이 가능하였고 인식 효율은 92-98%가 나왔다. 전립선비대증식에 대한 초음파영상을 가지고 컴퓨터영상처리분석을 제안하여 진단시 참고 자료가 될 것으로 기대한다.

• The Journal of Korean Institute of Communications and Information Sciences
• /
• v.22 no.7
• /
• pp.1487-1504
• /
• 1997

In this paper, the two-stage variable block-size multiresolution motion algorithm is proposed for an interframe coding scheme in the wavelet decomposition. An optimal bit allocagion between motion vectors and the prediction error in sense of minimizing the total bit rate is obtained by the proposed algorithm. The proposed algorithm consists of two stages for motion estimatation and only the first stage can be separated and run on its own. The first stage of the algorithm introduces a new method to give the lower bit rate of the displaced frame difference as well as a smooth motion field. In the second stage of the algorithm, the technique is introduced to have more accurate motion vectors in detailed areas, and to decrease the number of motion vectors in uniform areas. The algorithm aims at minimizin gthe total bit rate which is sum of the motion vectors and the displaced frame difference. The optimal bit allocation between motion vectors and displaced frame difference is accomplished by reducing the number of motion vectors in uniform areas and it is based on a botom-up construction of a quadtree. An entropy criterion aims at the control of merge operation. Simulation resuls show that the algorithm lends itself to the wavelet based image sequence coding and outperforms the conventional scheme by up to the maximum 0.28 bpp.

• The Journal of Korean Institute of Communications and Information Sciences
• /
• v.25 no.7A
• /
• pp.1017-1027
• /
• 2000

In this paper, we propose a new coding method based on the distance between vertex segments for vertex positions in the vertex-based shape coding. The pixel lines are divided into the segments of a fixed length, and the segments that have vertex pixels are called vertex segments. We analyze the probability distribution of the relative distance between vertex segments and prove that it depends only on the ratio between the number of vertices and the number of segments. Considering the coding efficiency and implementation complexity, we choose a particular ratio to make a code table. For each input image, the segment size is chosen according tothe ratio, and the relative segment distances are entropy coded. It is shown that the proposed method is efficient for the images with many vertices.

• Journal of the Institute of Convergence Signal Processing
• /
• v.5 no.2
• /
• pp.91-95
• /
• 2004

In this paper, we propose a fuzzy controller for the evaluation of the quantization parameters in the H.263 coder. Our method adopts the Mamdani method for fuzzification and adopts the centroid method for defuzzification respectively. The inputs are variance, entropy in the spatial domain, current motion vector and previous motion vector in the temporal. Fuzzy variables are determined to be compatible in visual characteristics and fuzzy membership function is induced and then, FAM banks are designed to reduce the number of rules. In this paper, fuzzy quantization has been applied to a practical video compression. This results show that the quality of decode image enhances and the rate control method using fuzzy quantization is effective.

• Journal of the Korean Society for Aeronautical & Space Sciences
• /
• v.49 no.10
• /
• pp.883-892
• /
• 2021

This paper provides a process for developing deep learning-based aerial object detection models that can run in realtime on onboard. To improve object detection performance, we pre-process and augment the training data in the training stage. In addition, we perform transfer learning and apply a weighted cross-entropy method to reduce the variations of detection performance for each class. To improve the inference speed, we have generated inference acceleration engines with quantization. Then, we analyze the real-time performance and detection performance on custom aerial image dataset to verify generalization.

• Journal of the Korea Institute of Military Science and Technology
• /
• v.22 no.2
• /
• pp.207-214
• /
• 2019

This paper proposes a new feature extraction method for automatically estimating the number of target and detecting the chaff using high range resolution profile(HRRP). Feature of one-dimensional range profile is expected to be limited or missing due to lack of information according to the time. The proposed method considers the dynamic movements of targets depending on the radial velocity. The observed HRRP sequence is used to construct a time-range distribution matrix, then assuming diverse radial velocities reflect the number of target and seduction chaff launch, the proposed method utilizes the characteristic of the gradient distribution on the time-range distribution matrix image, which is validated by electromagnetic computation data and dynamic simulation.

• Journal of Broadcast Engineering
• /
• v.26 no.1
• /
• pp.79-87
• /
• 2021

In this paper, we analysis the semi-supervised learning (SSL), which is adopted in order to train a deep learning-based classification model using the small number of labeled data. The conventional SSL techniques can be categorized into consistency regularization, entropy-based, and pseudo labeling. First, we describe the algorithm of each SSL technique. In the experimental results, we evaluate the classification accuracy of each SSL technique varying the number of labeled data. Finally, based on the experimental results, we describe the limitations of SSL technique, and suggest the research direction to improve the classification performance of SSL.

• International Journal of Computer Science & Network Security
• /
• v.22 no.1
• /
• pp.1-6
• /
• 2022

Exploration has always been an instinct of humans, and underwater life is as fascinating as it seems. So, for studying flora and fauna below water, there is a need for high-quality images. However, the underwater images tend to be of impaired quality due to various factors, which calls for improved and enhanced underwater images. There are various Histogram Equalization (HE) based techniques which could aid in solving these issues. Classifying the HE methods broadly, there is Global Histogram Equalization (GHE), Mean Brightness Preserving HE (MBPHE), Bin Modified HE (BMHE), and Local HE (LHE). Each of these HE extensions have their own pros and cons and thus, by considering them we have considered BBHE, CLAHE, BPDHE, BPDFHE, and DSIHE enhancement algorithms, which are based on Mean Brightness Preserving HE and Local HE, for this study. The performance is evaluated with non-reference performance measures like Entropy, UCIQE, UICM, and UIQM. In this study, we apply the enhancement algorithms on 300 images from the UIEB benchmark dataset and then apply the techniques of cascading fusion on the best-performing algorithms.