• 한국산림과학회지
• /
• 제96권5호
• /
• pp.561-566
• /
• 2007

This study was conducted to clarify runoff production processes in forested catchment through hydrograph separation using three-component mixing model based on the End Member Mixing Analysis (EMMA) model. The study area is located in the coniferous-forested experimental catchment, Gwangneung Gyeonggido near Seoul, Korea (N 37 45', E 127 09'). This catchment is covered by Pinus Korainensis and Abies holophylla planted at stocking rate of 3,000 trees $ha^{-1}$ in 1976. Thinning and pruning were carried out two times in the spring of 1996 and 2004 respectively. We monitored 8 successive events during the periods from June 15 to September 15, 2005. Throughfall, soil water and groundwater were sampled by the bulk sampler. Stream water was sampled every 2-hour through ISCO automatic sampler for 48 hours. The geochemical tracers were determined in the result of principal components analysis. The concentrations of $SO_4{^{2-}$ and $Na^+$ for stream water almost were distributed within the bivariate plot of the end members; throughfall, soil water and groundwater. Average contributions of throughfall, soil water and groundwater on producing stream flow for 8 events were 17%, 25% and 58% respectively. The amount of antecedent precipitation (AAP) plays an important role in determining which end members prevail during the event. It was found that ground water contributed more to produce storm runoff in the event of a small AAP compared with the event of a large AAP. On the other hand, rain water showed opposite tendency to ground water. Rain water in storm runoff may be produced by saturation overland flow occurring in the areas where soil moisture content is near saturation. AAP controls the producing mechanism for storm runoff whether surface or subsurface flow prevails.

• 한국BIM학회 논문집
• /
• 제6권2호
• /
• pp.29-38
• /
• 2016

Road bridges are deteriorating gradually, and it is forecasted that the number of road bridges aging over 30 years will increase by more than 3 times of the current number. To maintain road bridges in a safe condition, current safety conditions of the bridges must be estimated for repair or reinforcement. However, budget and professional manpower required to perform in-depth inspections of road bridges are limited. This study proposes an estimation model for safety rating of road bridges by analyzing the data from Facility Management System (FMS) and Yearbook of Road Bridges and Tunnel. These data include basic specifications, year of completion, traffic, safety rating, and others. The distribution of safety rating was imbalanced, indicating 91% of road bridges have safety ratings of A or B. To improve classification performance, five safety ratings were integrated into two classes of G (good, A and B) and P (poor ratings under C). This rearrangement was set because facilities with ratings under C are required to be repaired or reinforced to recover their original functionality. 70% of the original data were used as training data, while the other 30% were used for validation. Data of class P in the training data were oversampled by 3 times, and Repeated Incremental Pruning to Produce Error Reduction (RIPPER) algorithm was used to develop the estimation model. The results of estimation model showed overall accuracy of 84.8%, true positive rate of 67.3%, and 29 classification rule. Year of completion was identified as the most critical factor on affecting lower safety ratings of bridges.

• 한국지능시스템학회논문지
• /
• 제23권2호
• /
• pp.184-189
• /
• 2013

본 논문에서는 인터넷 및 인터넷 저장 공간에 제한없이 유통되고 있는 유해동영상을 필터링하기 위해 유해동영상에 포함된 특정 소리를 이용한 유해 동영상 필터링 시스템을 제안한다. 이를 위하여 소리의 특성을 잘 표현할 수 있는 Gaussian Mixture Model을 이용하였으며, 필터링 대상 데이터와 소리모델과의 유사도를 계산하기위해 프레임단위 유사도 추정을 이용하였다. 또, 실시간 처리를 위하여 비교대상 데이터의 수를 줄임으로서 실시간 처리가 가능한 프루닝 방법을 적용하였으며, 고정도의 구별 성능을 위하여 기존 화자식별에서 우수한 성능을 보였던 MWMR 방법을 적용하였다. 식별실험결과, 일반 영상과 유해 영상의 기준인 전체프레임 대비 유사도 높은 프레임의 비를 50%로 설정한 경우, 판별 오류율은 6.06%였으며, 프레임 비의 기준이 60%인 경우, 오류율은 3.03%를 나타내어 소리를 이용한 유해동영상 필터링 시스템이 효과적으로 일반영상과 유해영상을 구별할 수 있는 것을 확인하였다.

• 한국산림과학회지
• /
• 제95권5호
• /
• pp.556-561
• /
• 2006

This study was conducted to choose end-members and tracers for application of End Member Mixing Analysis (EMMA) model for the coniferous forest catchment, Gwangneung Gyeongi-do near Seoul metropolitan of South Korea (N $37^{\circ}$ 45', E $127^{\circ}$ 09'). This coniferous forest of Pinus Korainensis and Abies holophylla was planted at stocking rate of $3.0stems\;ha^{-1}$ in 1976. Thinning and pruning were carried out two times in the spring of 1996 and 2004 respectively. We monitored two successive rainfall events during ten days from June 26, 205 to July 5, 2005. Two storm events were selected to determine the end members and natural traces for hydrograph separation. The event 1 amounts to 161.9 m for two days from June 26 to 27, 2005. The event 2 precipitates to 139.2 mm for one day of July 1, 205. Throughfall, groundwater, soil water and stream water of the two events above were sampled through the bulk and automatic sampler. Their chemical properties were analyzed for prediction of the main tracer. The end members that contribute to the stream runoff were identified from the three components including groundwater, soil water and throughfall. Each component and stream water in the two events formed the suitable mixing diagram in case of chloride-nitrate ion and sulfate-potassium ion. Especially, chloride-nitrate ion was found to be the most suitable tracers for EMMA model in the two events.

• 방송공학회논문지
• /
• 제26권6호
• /
• pp.778-789
• /
• 2021

최근 다양한 분야에서 뛰어난 성능을 나타내는 Convolutional Neural Network(CNN)모델을 모바일 기기에서 사용하기 위한 다양한 연구가 진행되고 있다. 기존의 CNN 모델은 모바일 장비에서 사용하기에는 가중치의 크기가 크고 연산복잡도가 높다는 문제점이 있다. 이를 해결하기 위해 가중치의 표현 비트를 낮추는 가중치 양자화를 포함한 여러 경량화 방법들이 등장하였다. 많은 방법들이 다양한 모델에서 적은 정확도 손실과 높은 압축률을 나타냈지만, 대부분의 압축 모델들은 정확도 손실을 복구하기 위한 재학습 과정을 포함시켰다. 재학습 과정은 압축된 모델의 정확도 손실을 최소화하지만 많은 시간과 데이터를 필요로 하는 작업이다. Weight Quantization이후 각 층의 가중치는 정수형 행렬로 나타나는데 이는 이미지의 형태와 유사하다. 본 논문에서는 Weight Quantization이후 각 층의 정수 가중치 행렬을 이미지의 형태로 비디오 코덱을 사용하여 압축하는 방법을 제안한다. 제안하는 방법의 성능을 검증하기 위해 ImageNet과 Places365 데이터 셋으로 학습된 VGG16, Resnet50, Resnet18모델에 실험을 진행하였다. 그 결과 다양한 모델에서 2%이하의 정확도 손실과 높은 압축 효율을 달성했다. 또한, 재학습 과정을 제외한 압축방법인 No Fine-tuning Pruning(NFP)와 ThiNet과의 성능비교 결과 2배 이상의 압축효율이 있음을 검증했다.

• 한국지능시스템학회논문지
• /
• 제17권2호
• /
• pp.238-243
• /
• 2007

신경망 모델에 기반한 TAM 네트워크는 특별히 패턴분석에 효과적인 모델이다. TAM 네트워크는 입력층, 카테고리층, 출력층으로 구성되어 있다. 입력 및 출력 데이터에 대한 퍼지룰은 TAM 네트워크에서 얻어진다. 각 층에서 링크와 노드를 감소하기 위한 3가지의 프루닝룰을 사용하는 TAM 네크워크를 퍼지 TAM 네트워크라고 한다. 본 논문에서는 퍼지 TAM 네트워크를 조직리더에 대한 리더십 유형의 패턴분석에 적용하고 그 유용성을 보인다. 여기서, 입력층의 평가기준은 이고그램의 성격유형 관련변수의 값이고, 출력층의 목표값은 에니어그램의 성격유형과 관련된 리더십이다.

• 한국방송∙미디어공학회:학술대회논문집
• /
• 한국방송∙미디어공학회 2019년도 추계학술대회
• /
• pp.44-45
• /
• 2019

최근 CNN 등 인공신경망은 최근 이미지 분류, 객체 인식, 자연어 처리 등 다양한 분야에서 뛰어난 성능을 보이고 있다. 그러나, 대부분의 분야에서 보다 더 높은 성능을 얻기 위해 사용한 인공신경망 모델들은 파라미터 수 및 연산량 등이 방대하여, 모바일 및 IoT 디바이스 같은 연산량이나 메모리가 제한된 환경에서 추론하기에는 제한적이다. 따라서 연산량 및 모델 파라미터 수를 압축하기 위한 딥러닝 경량화 알고리즘이 연구되고 있다. 본 논문에서는 임베디트 보드에서의 압축된 CNN 모델의 성능을 검증한다. 인공지능 지원 맞춤형 칩인 QCS605 를 내장한 임베디드 보드에서 카메라로 입력한 영상에 대해서 원 CNN 모델과 압축된 CNN 모델의 분류 성능과 동작속도 비교 분석한다. 본 논문의 실험에서는 CNN 모델로 MobileNetV2, VGG16 을 사용했으며, 주어진 모델에서 가지치기(pruning) 기법, 양자화, 행렬 분해 등의 인공신경망 압축 기술을 적용하였을 때 원래의 모델 대비 추론 시간 및 분류의 정확도 성능을 분석하고 인공신경망 압축 기술의 유용성을 확인하였다.

• 한국경영과학회:학술대회논문집
• /
• 대한산업공학회/한국경영과학회 2006년도 춘계공동학술대회 논문집
• /
• pp.17-24
• /
• 2006

We consider the sensor location problem (SLP) on a given sensor field. We present the sensor field as grid of points. There are several types of sensors which have different detection ranges and costs. If a sensor is placed in some point, the points inside of its detection range can be covered. The coverage ratio decreases with distance. The problem we consider in this thesis is called multiple-type differential coverage sensor location problem (MDSLP). MDSLP is more realistic than SLP. The coverage quantities of points are different with their distance form sensor location in MDSLP. The objective of MDSLP is to minimize total sensor costs while covering every sensor field. This problem is known as NP-hard. We propose a new integer programming formulation of the problem. In comparison with the previous models, the new model has a smaller number of constraints and variables. This problem has symmetric structure in its solutions. This group is used for pruning in the branch-and-bound tree. We solved this problem by branch-and-cut(B&C) approach. We tested our algorithm on about 60 instances with varying sizes.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• 제4권6호
• /
• pp.1273-1293
• /
• 2010

Immune-inspired intrusion detection is a promising technology for network security, and well known for its diversity, adaptation, self-tolerance, etc. However, scalability and coverage are two major drawbacks of the immune-inspired intrusion detection systems (IIDSes). In this paper, we propose an IIDS framework, named GEP-IIDS, with improved basic system elements to address these two problems. First, an additional bio-inspired technique, gene expression programming (GEP), is introduced in detector (corresponding to detection rules) representation. In addition, inspired by the avidity model of immunology, new avidity/affinity functions taking the priority of attributes into account are given. Based on the above two improved elements, we also propose a novel immune algorithm that is capable of integrating two bio-inspired mechanisms (i.e., negative selection and positive selection) by using a balance factor. Finally, a pruning algorithm is given to reduce redundant detectors that consume footprint and detection time but do not contribute to improving performance. Our experimental results show the feasibility and effectiveness of our solution to handle the scalability and coverage problems of IIDS.

• Smart Structures and Systems
• /
• 제29권1호
• /
• pp.141-151
• /
• 2022

In recent years, the industry and research communities have focused on developing autonomous crack inspection approaches, which mainly include image acquisition and crack detection. In these approaches, mobile devices such as cameras, drones or smartphones are utilized as sensing platforms to acquire structural images, and the deep learning (DL)-based methods are being developed as important crack detection approaches. However, the process of image acquisition and collection is time-consuming, which delays the inspection. Also, the present mobile devices such as smartphones can be not only a sensing platform but also a computing platform that can be embedded with deep neural networks (DNNs) to conduct on-site crack detection. Due to the limited computing resources of mobile devices, the size of the DNNs should be reduced to improve the computational efficiency. In this study, an architecture called pruned crack recognition network (PCR-Net) was developed for the detection of structural cracks. A dataset containing 11000 images was established based on the raw images from bridge inspections. A pruning method was introduced to reduce the size of the base architecture for the optimization of the model size. Comparative studies were conducted with image processing techniques (IPTs) and other DNNs for the evaluation of the performance of the proposed PCR-Net. Furthermore, a modularly designed framework that integrated the PCR-Net was developed to realize a DL-based crack detection application for smartphones. Finally, on-site crack detection experiments were carried out to validate the performance of the developed system of smartphone-based detection of structural cracks.