• 한국항해항만학회지
• /
• 제36권7호
• /
• pp.553-559
• /
• 2012

It is extremely important to predict the logistics requirements in a scientific and rational way. However, in recent years, the improvement effect on the prediction method is not very significant and the traditional statistical prediction method has the defects of low precision and poor interpretation of the prediction model, which cannot only guarantee the generalization ability of the prediction model theoretically, but also cannot explain the models effectively. Therefore, in combination with the theories of the spatial economics, industrial economics, and neo-classical economics, taking city of Erdos as the research object, the study identifies the leading industry that can produce a large number of cargoes, and further predicts the static logistics generation of the Erdos and hinterlands. By integrating various factors that can affect the regional logistics requirements, this study established a logistics requirements potential model from the aspect of spatial economic principles, and expanded the way of logistics requirements prediction from the single statistical principles to an new area of special and regional economics.

• 한국멀티미디어학회논문지
• /
• 제24권5호
• /
• pp.659-666
• /
• 2021

With the continuous development of deep learning, human behavior recognition algorithms have achieved good results. However, in a multi-person recognition environment, the complex behavior environment poses a great challenge to the efficiency of recognition. To this end, this paper proposes a multi-person pose estimation model. First of all, the human detectors in the top-down framework mostly use the two-stage target detection model, which runs slow down. The single-stage YOLOv3 target detection model is used to effectively improve the running speed and the generalization of the model. Depth separable convolution, which further improves the speed of target detection and improves the model's ability to extract target proposed regions; Secondly, based on the feature pyramid network combined with context semantic information in the pose estimation model, the OHEM algorithm is used to solve difficult key point detection problems, and the accuracy of multi-person pose estimation is improved; Finally, the Euclidean distance is used to calculate the spatial distance between key points, to determine the similarity of postures in the frame, and to eliminate redundant postures.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• 제15권1호
• /
• pp.180-194
• /
• 2021

For small image registration, feature-based approaches are likely to fail as feature detectors cannot detect enough feature points from low-resolution images. The classic FFT approach's prediction accuracy is high, but the registration time can be relatively long, about several seconds to register one image pair. To achieve real-time and high-precision rigid registration for small images, we apply deep neural networks for supervised rigid transformation prediction, which directly predicts the transformation parameters. We train deep registration models with rigidly transformed CIFAR-10 images and STL-10 images, and evaluate the generalization ability of deep registration models with transformed CIFAR-10 images, STL-10 images, and randomly generated images. Experimental results show that the deep registration models we propose can achieve comparable accuracy to the classic FFT approach for small CIFAR-10 images (32×32) and our LSTM registration model takes less than 1ms to register one pair of images. For moderate size STL-10 images (96×96), FFT significantly outperforms deep registration models in terms of accuracy but is also considerably slower. Our results suggest that deep registration models have competitive advantages over conventional approaches, at least for small images.

• Smart Structures and Systems
• /
• 제30권3호
• /
• pp.303-315
• /
• 2022

The degradation of wheel tread may result in serious hazards in the railway operation system. Therefore, timely wheel defect diagnosis of in-service trains to avoid tragic events is of particular importance. The focus of this study is to develop a novel wheel defect detection approach based on the relevance vector machine (RVM) which enables online detection of potentially defective wheels with trackside monitoring data acquired under different running-speed conditions. With the dynamic strain responses collected by a trackside monitoring system, the cumulative Fourier amplitudes (CFA) characterizing the effect of individual wheels are extracted to formulate multiple probabilistic regression models (MPRMs) in terms of multi-kernel RVM, which accommodate both variables of vibration frequency and running speed. Compared with the general single-kernel RVM-based model, the proposed multi-kernel MPRM approach bears better local and global representation ability and generalization performance, which are prerequisite for reliable wheel defect detection by means of data acquired under different running-speed conditions. After formulating the MPRMs, we adopt a Bayesian null hypothesis indicator for wheel defect identification and quantification, and the proposed method is demonstrated by utilizing real-world monitoring data acquired by an FBG-based trackside monitoring system deployed on a high-speed trial railway. The results testify the validity of the proposed method for wheel defect detection under different running-speed conditions.

• Smart Structures and Systems
• /
• 제29권1호
• /
• pp.153-166
• /
• 2022

Identifying fine cracks in steel bridge facilities is a challenging task of structural health monitoring (SHM). This study proposed an end-to-end crack image segmentation framework based on a one-step Convolutional Neural Network (CNN) for pixel-level object recognition with high accuracy. To particularly address the challenges arising from small object detection in complex background, efforts were made in loss function selection aiming at sample imbalance and module modification in order to improve the generalization ability on complicated images. Specifically, loss functions were compared among alternatives including the Binary Cross Entropy (BCE), Focal, Tversky and Dice loss, with the last three specialized for biased sample distribution. Structural modifications with dilated convolution, Spatial Pyramid Pooling (SPP) and Feature Pyramid Network (FPN) were also performed to form a new backbone termed CrackDet. Models of various loss functions and feature extraction modules were trained on crack images and tested on full-scale images collected on steel box girders. The CNN model incorporated the classic U-Net as its backbone, and Dice loss as its loss function achieved the highest mean Intersection-over-Union (mIoU) of 0.7571 on full-scale pictures. In contrast, the best performance on cropped crack images was achieved by integrating CrackDet with Dice loss at a mIoU of 0.7670.

• 한국측량학회지
• /
• 제36권6호
• /
• pp.469-481
• /
• 2018

딥러닝은 인간의 학습 및 인지능력을 닮은 인공지능을 실현하기 위해 여러 분야에서 활용하고 있으며, 높은 사양의 컴퓨팅 파워가 요구되고 연산 시간이 많이 소요되는 복잡한 구조의 인공신경망에 의한 딥러닝은 컴퓨터 사양이 향상됨에 따라 성능이 개선된 다양한 딥러닝 모델이 개발되고 있다. 본 논문의 주요 목적은 영상의 딥러닝을 위한 합성곱 신경망 중에서 최근에 FAIR (Facebook AI Research)에서 개발한 Mask R-CNN을 이용하여 항공영상에서 건물을 탐지하고 성능을 평가하는 것이다. Mask R-CNN은 영역기반의 합성곱 신경망으로서 픽셀 정확도까지 객체를 의미적으로 분할하기 위한 딥러닝 모델로서 성능이 가장 우수한 것으로 평가받고 있다. 딥러닝 모델의 성능은 신경망 구조뿐 아니라 학습 능력에 의해 결정된다. 이를 위해 본 논문에서는 모델의 학습에 이용한 영상에 다양한 변화를 주어 학습 능력을 분석하였으며, 딥러닝의 궁극적 목표인 범용화의 가능성을 평가하였다. 향후 연구방안으로는 영상에만 의존하지 않고 다양한 공간정보 데이터를 복합적으로 딥러닝 모델의 학습에 이용하여 딥러닝의 신뢰성과 범용화가 향상될 것으로 판단된다.

• 한국정보통신학회:학술대회논문집
• /
• 한국정보통신학회 2014년도 추계학술대회
• /
• pp.733-735
• /
• 2014

표본기반 초해상도(Super Resolution 이하 SR) 기법은 데이터베이스에 저장된 고해상도 영상의 패치와 저해상도 영상의 패치 사이에 대응관계를 이용하여, 저해상도의 입력영상에 가장 유사한 고해상도 패치를 덧붙여서 고해상도를 구성하는 방식이다. 이러한 방식은 한 장의 영상만으로 고해상도 영상을 얻을 수 있고, 위의 과정을 반복하여 2배 이상의 확대된 영상을 얻을 수 있어서 기존의 고전적 SR의 문제점을 해결할 수 있다. 표본기반 SR의 방법들 중 네이버 임베딩(Neighbor Embedding 이하 NE) 기법의 기본 원리는 지역적 선형 임베딩이라는 매니폴드 학습방법의 개념과 같다. 그러나 네이버 임베딩의 빈약한 일반화 능력으로 인하여 알고리즘의 성능을 크게 저하시킨다. 이유는 국부학습 데이터 집합의 크기가 너무 작아서 NE 알고리즘의 성능을 현저히 저하시킨다. 본 논문에서는 이와 같은 문제점을 해결하기 위해서 일반화 능력이 뛰어난 Support Vector Regression(이하 SVR)기반 개선된 NE를 제안하였다. 저해상도 입력 패치가 주어지면 SVR 기반 개선된 NE를 이용하여 고해상도의 해당 화소 값을 예측하였다. 실험을 통하여 제안된 기법이 기존의 보간법 및 NE 기법 등에 비해 정량적인 척도 및 시각적으로 향상된 결과를 보여 주었다.

• Journal of Electrical Engineering and Technology
• /
• 제12권2호
• /
• pp.830-839
• /
• 2017

Support vector machine (SVM) is introduced as an effective fault diagnosis technique based on dissolved gases analysis (DGA) for oil-immersed transformers with maximum generalization ability; however, the applicability of the SVM is highly affected due to the difficulty of selecting the SVM parameters appropriately. Therefore, a novel approach combing SVM with improved imperialist competitive algorithm (IICA) for fault diagnosis of oil-immersed transformers was proposed in the paper. The improved ICA, which is proved to be an effective optimization approach, is employed to optimize the parameters of SVM. Cross validation and normalizations were applied in the training processes of SVM and the trained SVM model with the optimized parameters was established for fault diagnosis of oil-immersed transformers. Three classification benchmark sets were studied based on particle swarm optimization SVM (PSOSVM) and IICASVM with four multiple classification schemes to select the best scheme for transformer fault diagnosis. The results show that the proposed model can obtain higher diagnosis accuracy than other methods. The comparisons confirm that the proposed model is an effective approach for classification problems.

• 전자공학회논문지
• /
• 제51권2호
• /
• pp.148-155
• /
• 2014

본 논문에서는 파형 신호에서 패치를 추출하고 이를 패치 그래프로 구성한 다음, 이로부터 대표적인 다양체 임베딩 방식인 컴뮤트 타임 임베딩 기법을 구현하고, 이의 특성을 분석한다. 특히 음성 신호나 악기 음 등, 시간에 따라 스펙트럼이 가변적인 신호를 임베딩하면 스펙트럼의 변화에도 불구하고 그 신호 고유의 기하 구조를 생섬함을 실험으로 확인한다. 다양체 임베딩은 비선형 공간에 놓여 있는 고차원 데이터를 저차원 공간으로의 효율적인 맵을 가능하게 하지만 그래프 구성에 이용된 데이터에 대한 정보만 알 수 있고 그렇지 않은 데이터(out-of-sample data)에 대해서는 정보를 얻기 어렵다. 따라서 다양체 임베딩은 데이터 클러스터링에 적절히 적용 가능하지만, 훈련 과정을 통해 얻은 정보를 기초로 유추 기능이 요구되는 인식 등에는 응용하기 어려운 제약이 따른다. 이를 해결하기 위하여 본 논문에서는 다양체 임베딩이 인식 분야에도 적용 가능하도록 새로운 알고리즘을 제안하고 악기 음 분류 실험을 통하여 그 특성을 분석한다.

• 한국전문물리치료학회지
• /
• 제20권3호
• /
• pp.9-18
• /
• 2013

The study aimed to compare the effect of the treadmill walking training combined with obstacle-crossing (TWT-OC) and treadmill walking training (TWT) on the walking function of patients with chronic stroke. 29 subjects volunteered to participate in this study; they were randomly assigned to either TWT-OC group (15 subjects) or TWT group (14 subjects). Subjects from the TWT-OC group underwent a treadmill walking combined with obstacles-crossing for 30 minutes daily, three days a week for four weeks, whereas subjects from the TWT group received only a treadmill walking. The 10 m walk test (10MWT), 6-min walk test (6MWT), berg balance scale (BBS), timed up and go test (TUG), activities-specific balance confidence-Korean version (ABC-K), and walking ability questionnaire (WAQ) were measured before and after the 4-week training. The TWT-OC group showed significantly better functional mobility of walking and balance measured by 6MWT (p<.01), BBS (p<.01), and TUG scores (p<.05) when compared with those of the TWT group. Further, within-group comparison showed significant improvement in all variables (p<.01) except for 10MWT. These findings suggest that the TWT-OC and TWT may be helpful for improving the walking function of patients with chronic stroke, and the TWT-OC has probably more favorable outcomes for chronic stroke, however, further trials with wider range of subjects are warranted for generalization and clinical relevance.