• Journal of Korea Multimedia Society
• /
• v.23 no.12
• /
• pp.1540-1551
• /
• 2020

Aiming at the problem that the existing human behavior recognition algorithm cannot fully utilize the multi-level spatio-temporal information of the network, a human behavior recognition algorithm based on a dense three-dimensional residual network is proposed. First, the proposed algorithm uses a dense block of three-dimensional residuals as the basic module of the network. The module extracts the hierarchical features of human behavior through densely connected convolutional layers; Secondly, the local feature aggregation adaptive method is used to learn the local dense features of human behavior; Then, the residual connection module is applied to promote the flow of feature information and reduced the difficulty of training; Finally, the multi-layer local feature extraction of the network is realized by cascading multiple three-dimensional residual dense blocks, and use the global feature aggregation adaptive method to learn the features of all network layers to realize human behavior recognition. A large number of experimental results on benchmark datasets KTH show that the recognition rate (top-l accuracy) of the proposed algorithm reaches 93.52%. Compared with the three-dimensional convolutional neural network (C3D) algorithm, it has improved by 3.93 percentage points. The proposed algorithm framework has good robustness and transfer learning ability, and can effectively handle a variety of video behavior recognition tasks.

• Journal of the Korean Society of Surveying, Geodesy, Photogrammetry and Cartography
• /
• v.37 no.1
• /
• pp.19-28
• /
• 2019

Images captured by drone have the advantage of quickly constructing spatial information in small areas and are applied to fields that require quick decision making. If an image registration technique that can automatically register the drone image on the ortho-image with the ground coordinate system is applied, it can be used for various analyses. In this study, a methodology for geo-referencing of a single image and sequential images using drones was proposed even if they differ in spatio-temporal resolution using linear features and distinct points. Through the method using linear features, projective transformation parameters for the initial geo-referencing between images were determined, and then finally the geo-referencing of the image was performed through the template matching for distinct points that can be extracted from the images. Experimental results showed that the accuracy of the geo-referencing was high in an area where relief displacement of the terrain was not large. On the other hand, there were some errors in the quantitative aspect of the area where the change of the terrain was large. However, it was considered that the results of geo-referencing of the sequential images could be fully utilized for the qualitative analysis.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.16 no.12
• /
• pp.3991-4007
• /
• 2022

In the public healthcare, a computational system that can automatically and efficiently detect and classify falls from a video sequence has significant potential. With the advancement of deep learning, which can extract temporal and spatial information, has become more widespread. However, traditional 3D CNNs that usually adopt shallow networks cannot obtain higher recognition accuracy than deeper networks. Additionally, some experiences of neural network show that the problem of gradient explosions occurs with increasing the network layers. As a result, an enhanced three-dimensional ResNet-based method for fall detection (3D-ERes-FD) is proposed to directly extract spatio-temporal features to address these issues. In our method, a 50-layer 3D residual network is used to deepen the network for improving fall recognition accuracy. Furthermore, enhanced residual units with four convolutional layers are developed to efficiently reduce the number of parameters and increase the depth of the network. According to the experimental results, the proposed method outperformed several state-of-the-art methods.

• Journal of Korea Multimedia Society
• /
• v.20 no.10
• /
• pp.1637-1645
• /
• 2017

Target/Nontarget classification can be divided into the study of shape estimation of the target analysing reflected echo signal and of type classification of the target using acoustical features. In active sonar system, the feature vectors are extracted from the signal reflected from the target, and an classification algorithm is applied to determine whether the received signal is a target or not. However, received sonar signals can be distorted in the underwater environments, and the spatio-temporal characteristics of active sonar signals change according to the aspect of the target. In addition, it is very difficult to collect real sea-trial data for research. In this paper, target/non-target classification were performed using real sea-trial data. Feature vectors are extracted using MFCC(Mel-Frequency Cepstral Coefficients), filterbank energy in the Fourier spectrum and wavelet domain. For the performance verification, classification experiments were performed using backpropagation neural network classifiers.

• Proceedings of the Korean Information Science Society Conference
• /
• 2005.11b
• /
• pp.892-894
• /
• 2005

본 논문에서는 스테레오 영상에서 깊이 정보를 추출하여 사람의 자세를 학습된 2차원 깊이 영상들의 선형 결함으로 표현하여 3차원 인체 모델을 재구성하는 방법을 제안한다. 한 장의 2차원 깊이 영상으로 최소 제곱법을 이용하여 프로토타입 깊이 영상의 선형 결합으로 표현되는 최적의 계수를 찾을 수 있다. 입력된 깊이 영상의 3차원 인체 모델은 프로토타입 깊이 영상에서 예측된 계수를 적용하여 생성한다. 학습 단계에서는 데이터를 계층적으로 나누어 모델을 생성한다. 또한, 재구성 단계에서는 실루엣 영상과 깊이 영상으로부터 계층적으로 나누어진 학습 데이터를 이용하여 3차원 인체 자세를 재구성한다. 학습 및 재구성의 마지막 단계에서는 실루엣 영상 대신 깊이 영상을 이용하여 3차원 인체 모델을 재구성한다. 한 장의 실루엣 영상을 이용하면 영상의 노이즈에 민감하기 때문에 재구성 단계의 상위 레벨에서는 실루엣 영상의 누적 영상을 이용한다. 실험 결과는 제안된 방법이 효율적으로 3차원 인체 자세를 재구성함을 보여준다.

• Proceedings of the Korean Association of Geographic Inforamtion Studies Conference
• /
• 2008.06a
• /
• pp.59-66
• /
• 2008

최근 유비쿼터스 컴퓨팅 환경이 발전함에 따라 교통(u-Transport), 복지(u-Care), 문화(u-Fun), 환경(u-Green), 산업(u-Business), 행정(u-Government), 도시(u-City) 뿐만 아니라 사용자의 위치와 다양한 공간 정보를 제공하는 u-GIS가 유비쿼터스 컴퓨팅 환경의 핵심 요소 기술로 대두되고 있다. 이에 본 논문에서는 기존의 PC용 MMDBMS인 HS QLDB를 확장하여 모바일 장치에서 시공간 데이타를 효율적으로 처리 및 관리할 수 있는 내장형 시공간 DBMS를 설계 및 구현하였다. 내장형 시공간 DBMS는 OpenGIS "Simple Features Specification for SQL"에서 명시하는 공간 데이타 타입과 공간 연산자를 확장하여 시공간 데이타 타입과 시공간 연산자를 제공하며, 시공간 데이타 특성들 고려한 산술 연산 코딩 압축 기법을 제공하고, 모바일 저장 장치인 플래쉬 메모리에서 효율적인 시공간 데이타 검색을 위한 시공간 인덱스를 지원한다. 그리고, 내장형 시공간 DBMS와 U-GIS 서버 사이에서 시공간 데이타 수입/수출의 성능 향상을 위한 데이타 캐슁 기능과 DBMS의 안정성을 위한 백업/복구 기능을 지원한다.

• KIPS Transactions on Software and Data Engineering
• /
• v.2 no.3
• /
• pp.209-218
• /
• 2013

Early researches in human action recognition have focused on tracking and classifying articulated body motions. Such methods required accurate segmentation of body parts, which is a sticky task, particularly under realistic imaging conditions. Recent trends of work have become popular towards the use of more and low-level appearance features such as spatio-temporal interest points. Given the great progress in pose estimation over the past few years, redefined views about pose-based approach are needed. This paper addresses the issues of whether it is sufficient to train a classifier only on low-level appearance features in appearance approach and proposes effective pose-based approach with pose estimation for emotional action recognition. In order for these questions to be solved, we compare the performance of pose-based, appearance-based and its combination-based features respectively with respect to scenario of various emotional action recognition. The experiment results show that pose-based features outperform low-level appearance-based approach of features, even when heavily spoiled by noise, suggesting that pose-based approach with pose estimation is beneficial for the emotional action recognition.

• Journal of Korea Water Resources Association
• /
• v.56 no.4
• /
• pp.235-243
• /
• 2023

Due to the revision of the River Act and the enactment of the Act on the Investigation, Planning, and Management of Water Resources, a regular bed change survey has become mandatory and a system is being prepared such that local governments can manage water resources in a planned manner. Since the topography of a bed cannot be measured directly, it is indirectly measured via contact-type depth measurements such as level survey or using an echo sounder, which features a low spatial resolution and does not allow continuous surveying owing to constraints in data acquisition. Therefore, a depth measurement method using remote sensing-LiDAR or hyperspectral imaging-has recently been developed, which allows a wider area survey than the contact-type method as it acquires hyperspectral images from a lightweight hyperspectral sensor mounted on a frequently operating drone and by applying the optimal bandwidth ratio search algorithm to estimate the depth. In the existing hyperspectral remote sensing technique, specific physical quantities are analyzed after matching the hyperspectral image acquired by the drone's path to the image of a surface unit. Previous studies focus primarily on the application of this technology to measure the bathymetry of sandy rivers, whereas bed materials are rarely evaluated. In this study, the existing hyperspectral image-based water depth estimation technique is applied to rivers with vegetation, whereas spatio-temporal hyperspectral imaging and cross-sectional hyperspectral imaging are performed for two cases in the same area before and after vegetation is removed. The result shows that the water depth estimation in the absence of vegetation is more accurate, and in the presence of vegetation, the water depth is estimated by recognizing the height of vegetation as the bottom. In addition, highly accurate water depth estimation is achieved not only in conventional cross-sectional hyperspectral imaging, but also in spatio-temporal hyperspectral imaging. As such, the possibility of monitoring bed fluctuations (water depth fluctuation) using spatio-temporal hyperspectral imaging is confirmed.

• Journal of the Korea Academia-Industrial cooperation Society
• /
• v.12 no.1
• /
• pp.507-512
• /
• 2011

Recently, various new devices have been introduced, which are capable of quickly measuring river hydrodynamic and morphologic features in the very broad riverine area. These devices are changing paradigm of understanding river characteristics in terms of data-driven aspect rather than the conventional numerical modeling approaches based on the limited field observations. This paper demonstrates the representative features and applications of the several recent riverine devices such as ADCP, LSPIV, MBES and ABL. In addition, the paper introduces an example of river information system that incorporates and relates such two- and three-dimensional hydrodynamic and morphologic data on top of geographic information system, where their spatio-temporal variations are also able to be tracked.

• The Journal of the Acoustical Society of Korea
• /
• v.40 no.3
• /
• pp.228-233
• /
• 2021

In this paper, we propose a method to extract and classify features of multi-channel ElectroEncephalo Graphy (EEG) that change according to various musical tempo stimuli. In the proposed method, a 3D convolutional bidirectional gated recurrent neural network extracts spatio-temporal and long time-dependent features from the 3D EEG input representation transformed through the preprocessing. The experimental results show that the proposed tempo stimuli classification method is superior to the existing method and the possibility of constructing a music-based brain-computer interface.