• Journal of the Korean Institute of Intelligent Systems
• /
• v.17 no.7
• /
• pp.924-929
• /
• 2007

Captions which appear in images include information that relates to the images. In order to obtain the information carried by captions, the methods for text extraction from images have been developed. However, most existing methods can be applied to captions with fixed height of stroke's width. We propose a method which can be applied to various caption size. Our method is based on connected components. And then the edge pixels are detected and grouped into connected components. We analyze the properties of connected components and build a neural network which discriminates connected components which include captions from ones which do not. Experimental data is collected from broadcast programs such as news, documentaries, and show programs which include various height caption. Experimental result is evaluated by two criteria : recall and precision. Recall is the ratio of the identified captions in all the captions in images and the precision is the ratio of the captions in the objects identified as captions. The experiment shows that the proposed method can efficiently extract captions various in size.

• Journal of Broadcast Engineering
• /
• v.26 no.4
• /
• pp.419-428
• /
• 2021

Shadow removal plays a key role for the pre-processing of image processing techniques such as object tracking and detection. With the advances of image recognition based on deep convolution neural networks, researches for shadow removal have been actively conducted. In this paper, we propose a novel method for shadow removal, which utilizes self attention distillation to extract semantic features. The proposed method gradually refines results of shadow detection, which are extracted from each layer of the proposed network, via top-down distillation. Specifically, the training procedure can be efficiently performed by learning the contextual information for shadow removal without shadow masks. Experimental results on various datasets show the effectiveness of the proposed method for shadow removal under real world environments.

• Journal of Broadcast Engineering
• /
• v.26 no.6
• /
• pp.725-737
• /
• 2021

Although the performance of cameras is gradually improving now, there are noise in the acquired digital images from the camera, which acts as an obstacle to obtaining high-resolution images. Traditionally, a filtering method has been used for denoising, and a convolutional neural network (CNN), one of the deep learning techniques, has been showing better performance than traditional methods in the field of image denoising, but the details in images could be lost during the learning process. In this paper, we present a CNN for image denoising, which improves image details by learning the details of the image based on wavelet transform. The proposed network uses two subnetworks for detail enhancement and noise extraction. The experiment was conducted through Gaussian noise and real-world noise, we confirmed that our proposed method was able to solve the detail loss problem more effectively than conventional algorithms, and we verified that both objective quality evaluation and subjective quality comparison showed excellent results.

• Journal of the Korea Society of Computer and Information
• /
• v.27 no.8
• /
• pp.69-75
• /
• 2022

In this paper, a method for predicting the speed for each lane from the link speed using an artificial neural network is presented to increase the accuracy of predicting the required time of a driving route. The time required for passing through a link is observed differently depending on the direction of going straight, turning right, or turning left at the intersection of the end of the link. Therefore, it is necessary to predict the speed according to the vehicle's traveling direction. Data required for learning and verification were constructed by refining the data measured at the Gongpyeong intersection of Gukchaebosang-ro in Daegu Metropolitan City and four adjacent intersections around it. Five neural network models were used. In addition, error analysis of the prediction was performed to select a neural network experimentally suitable for the research purpose. Experimental results showed that the error in the estimation of the time required for each lane decreased by 17.4% for the straight lane, 4.4% for the right-turn lane, and 3.9% for the left-turn lane. This experiment is the result of analyzing only one link. If the entire pathway is tested, the effect is expected to be greater.

• Proceedings of the Korean Institute of Information and Commucation Sciences Conference
• /
• 2021.05a
• /
• pp.43-45
• /
• 2021

The short-term quantitative precipitation prediction (QPF) system is important socially and economically to prevent damage from severe weather. Recently, many studies for short-term QPF model applying the Deep Neural Network (DNN) has been conducted. These studies require the sophisticated pre-processing because the mistreatment of various and vast meteorological data sets leads to lower performance of QPF. Especially, for more accurate prediction of the non-linear trends in precipitation, the dataset needs to be carefully handled based on the physical and dynamical understands the data. Thereby, this paper proposes the following approaches: i) refining and combining major factors (weather radar, terrain, air temperature, and so on) related to precipitation development in order to construct training data for pattern analysis of precipitation; ii) producing predicted precipitation fields based on Convolutional with ConvLSTM. The proposed algorithm was evaluated by rainfall events in 2020. It is outperformed in the magnitude and strength of precipitation, and clearly predicted non-linear pattern of precipitation. The algorithm can be useful as a forecasting tool for preventing severe weather.

• Journal of Intelligence and Information Systems
• /
• v.15 no.4
• /
• pp.123-140
• /
• 2009

Fuzzy neural network is an integrated model of artificial neural network and fuzzy system and it has been successfully applied in control and forecasting area. Recently ANFIS(Adaptive Network-based Fuzzy Inference System) has been noticed widely among various fuzzy neural network models because of outstanding performance of control and forecasting accuracy. ANFIS has capability to refine its fuzzy rules interactively with human expert. In particular, when we use initial rule structure for machine learning which is generated from human expert, it is highly probable to reach global optimum solution as well as shorten time to convergence. We propose metrics to evaluate interpretability of generated rules as a means of acquiring domain knowledge and compare level of interpretability of ANFIS fuzzy rules to those of C5.0 classification rules. The proposed metrics also can be used to evaluate capability of rule generation for the various machine learning methods.

• The Transactions of the Korean Institute of Electrical Engineers D
• /
• v.53 no.12
• /
• pp.804-810
• /
• 2004

There are two important things for the general purpose neural network processor. The first is a capability to build various structures of neural network, and the second is to be able to support suitable learning method for that neural network. Some way to process various learning algorithms is required for on-chip learning, because the more neural network types are to be handled, the more learning methods need to be built into. In this paper, an improved hardware structure is proposed to compute various kinds of learning algorithms flexibly. The hardware structure is based on the existing modular neural network structure. It doesn't need to add a new circuit or a new program for the learning process. It is shown that rearrangements of the existing processing elements can produce several neural network learning modules. The performance and utilization of this module are analyzed by comparing with other neural network chips.

• Proceedings of the Korean Society of Broadcast Engineers Conference
• /
• 2017.11a
• /
• pp.193-196
• /
• 2017

본 논문에서는 콘트라스트 복원 기법을 통한 효과적인 비 제거 알고리즘을 제안한다. 기존의 비 제거 알고리즘은 영상에 나타난 비를 제거하는 과정에서 빗줄기로 인한 콘트라스트 감소 현상을 제거하지 못하는 단점이 존재한다. 이에 본 논문에서는 합성곱 신경망을 이용하여 빗줄기를 제거하는 기존의 알고리즘에 덧붙여 빗줄기로 인해 발생된 콘트라스트 변화량을 추정하여 복원하는 과정을 추가함으로써 기존의 알고리즘보다 효과적으로 비를 제거하는 방법을 제시한다. 실험을 통해 성능을 분석한 결과, 제안하는 알고리즘은 모든 결과 영상에 대하여 기존의 알고리즘에 비해 높은 화질 평가 수치를 나타내었으며 주관적으로도 더 선명한 결과를 보여주었다.

• Proceedings of the Korean Society of Broadcast Engineers Conference
• /
• fall
• /
• pp.113-116
• /
• 2021

일반적으로 딥러닝 기반의 객체 검출(Object Detection)기법은 합성곱 신경망(Convolutional Neural Network, CNN)을 통해 입력된 영상의 특징(Feature)을 추출하여 이를 통해 객체 검출을 수행한다. 최근 자연어 처리 분야에서 획기적인 성능을 보인 트랜스포머(Transformer)가 영상 분류, 객체 검출과 같은 컴퓨터 비전 작업을 수행하는데 있어 경쟁력이 있음이 드러나고 있다. 본 논문에서는 YOLOv4-CSP의 CSP 블록을 개선한 one-stage 방식의 객체 검출 네트워크를 제안한다. 개선된 CSP 블록은 트랜스포머(Transformer)의 멀티 헤드 어텐션(Multi-Head Attention)과 CSP 형태의 공간 피라미드 풀링(Spatial Pyramid Pooling, SPP) 연산을 기반으로 네트워크의 Backbone과 Neck에서의 feature 학습을 돕는다. 본 실험은 MSCOCO test-dev2017 데이터 셋으로 평가하였으며 제안하는 네트워크는 YOLOv4-CSP의 경량화 모델인 YOLOv4s-mish에 대하여 평균 정밀도(Average Precision, AP)기준 2.7% 향상된 검출 정확도를 보인다.

• Journal of Internet Computing and Services
• /
• v.21 no.5
• /
• pp.119-127
• /
• 2020

As the electric vehicle (EV) market in South Korea grows, it is required to expand charging facilities to respond to rapidly increasing EV charging demand. In order to conduct a comprehensive facility planning, it is necessary to forecast future demand for electricity and systematically analyze the impact on the load capacity of facilities based on this. In this paper, we design and develop a Long Short-Term Memory (LSTM) neural network model that predicts the daily peak electric load at each charging station using the EV charging data of KEPCO. First, we obtain refined data through data preprocessing and outlier removal. Next, our model is trained by extracting daily features per charging station and constructing a training set. Finally, our model is verified through performance analysis using a test set for each charging station type, and the limitations of our model are discussed.