• Journal of Information Processing Systems
• /
• 제17권2호
• /
• pp.411-425
• /
• 2021

In this paper, a vehicle recognition algorithm based on deep convolutional neural network and compression dictionary is proposed. Firstly, the network structure of fine vehicle recognition based on convolutional neural network is introduced. Then, a vehicle recognition system based on multi-scale pyramid convolutional neural network is constructed. The contribution of different networks to the recognition results is adjusted by the adaptive fusion method that adjusts the network according to the recognition accuracy of a single network. The proportion of output in the network output of the entire multiscale network. Then, the compressed dictionary learning and the data dimension reduction are carried out using the effective block structure method combined with very sparse random projection matrix, which solves the computational complexity caused by high-dimensional features and shortens the dictionary learning time. Finally, the sparse representation classification method is used to realize vehicle type recognition. The experimental results show that the detection effect of the proposed algorithm is stable in sunny, cloudy and rainy weather, and it has strong adaptability to typical application scenarios such as occlusion and blurring, with an average recognition rate of more than 95%.

• 디지털융복합연구
• /
• 제11권12호
• /
• pp.345-352
• /
• 2013

본 논문에서는 GPU(Graphics Processing Unit)에서 데이터를 처리할 수 있게 하여 단일 영상에서 효율적으로 깊이를 추정하는 방법을 제안한다. 단일 영상은 카메라의 투영 과정에 의해 깊이 정보가 소실되게 되며 영상에서 소실된 깊이를 추정하기 위해서 단안 단서를 이용한다. 제안하는 깊이 추정 알고리즘은 좀 더 신뢰성 있는 깊이를 추정하고자 여러 단안 단서를 이용하며 에너지 최소화를 통해 단안 단서들을 결합한다. 그러나 여러 단안 단서들을 고려해야하기 때문에 처리해야 할 데이터가 많은 단점이 존재한다. 따라서 GPGPU(General Purpose Graphics Processing Unit)를 통해 데이터를 병렬적으로 처리하게 하여 효율적으로 깊이를 추정하는 방법을 제안한다. 객관적인 효율성을 검증하기 위해 PSNR(Peak Signal to Noise Ratio)을 통해 실험하였으며 GPGPU을 이용함으로써 알고리즘의 수행시간을 평균 61.22% 감소시켰다.

• 콘크리트학회논문집
• /
• 제17권3호
• /
• pp.369-374
• /
• 2005

이 연구의 목적은 화상처리 기법과 신경회로망을 이용하여 다섯가지 균열 패턴 즉, 횡방향, 종방향, 대각선($-45^{\circ}$) 대각선($+45^{\circ}$) 그리고 비방향성 균열의 패턴을 인식할 수 있는 기법을 제안하는 것이다. 제안된 화상처리 알고리즘과 인공 신경회로망 모델은 MATLAB 언어를 이용하여 구현하였다. 인공 신경회로망의 입력층에 들어갈 패턴인자는 Total projection technique를 통해 구하였으며, 인공 신경회로망의 구조(은닉층의 수와 은닉노드의 수)와 가중치 값은 가상 균열 화상을 사용하여 학습을 통해 결정하였다. 인공 신경회로망의 학습은 Bayesian regularization 기법을 도입함으로써 과적합 문제가 발생하지 않도록 하였으며, 이 연구에서 제안한 기법의 적합성을 판정하기 위하여 총 38개의 실제 균열 화상을 사용하여 시험하였다. 검증 시험 결과내에서는 이 연구에서 제안한 기법이 사람의 균열 패턴 인식결과와 정확히 일치하는 결과것으로 나타났다.

• 대기
• /
• 제22권2호
• /
• pp.149-161
• /
• 2012

Going by the research results of the past, of all the uncertainties resulting from the research on climate change, the uncertainty caused by the climate change scenario has the highest degree of uncertainty. Therefore, depending upon what kind of climate change scenario one adopts, the projection of the water resources in the future will differ significantly. As a matter of principle, it is highly recommended to utilize all the GCM scenarios offered by the IPCC. However, this could be considered to be an impractical alternative if a decision has to be made at an action officer's level. Hence, as an alternative, it is deemed necessary to select several scenarios so as to express the possible number of cases to the maximum extent possible. The objective standards in selecting the climate change scenarios have not been properly established and the scenarios have been selected, either at random or subject to the researcher's discretion. In this research, a new scenario selection process, in which it is possible to have the effect of having utilized all the possible scenarios, with using only a few principal scenarios and maintaining some of the uncertainties, has been suggested. In this research, the use of cluster analysis and the selection of a representative scenario in each cluster have efficiently reduced the number of climate change scenarios. In the cluster analysis method, the K-means clustering method, which takes advantage of the statistical features of scenarios has been employed; in the selection of a representative scenario in each cluster, the selection method was analyzed and reviewed and the PDF method was used to select the best scenarios with the closest simulation accuracy and the principal scenarios that is suggested by this research. In the selection of the best scenarios, it has been shown that the GCM scenario which demonstrated high level of simulation accuracy in the past need not necessarily demonstrate the similarly high level of simulation accuracy in the future and various GCM scenarios were selected for the principal scenarios. Secondly, the "Maximum entropy" which can quantify the uncertainties of the climate change scenario has been used to both quantify and compare the uncertainties associated with all the scenarios, best scenarios and the principal scenarios. Comparison has shown that the principal scenarios do maintain and are able to better explain the uncertainties of all the scenarios than the best scenarios. Therefore, through the scenario selection process, it has been proven that the principal scenarios have the effect of having utilized all the scenarios and retaining the uncertainties associated with the climate change to the maximum extent possible, while reducing the number of scenarios at the same time. Lastly, the climate change scenario most suitable for the climate on the Korean peninsula has been suggested. Through the scenario selection process, of all the scenarios found in the 4th IPCC report, principal climate change scenarios, which are suitable for the Korean peninsula and maintain most of the uncertainties, have been suggested. Therefore, it is assessed that the use of the scenario most suitable for the future projection of water resources on the Korean peninsula will be able to provide the projection of the water resources management that maintains more than 70~80% level of uncertainties of all the scenarios.

• 한국광학회지
• /
• 제33권6호
• /
• pp.267-274
• /
• 2022

쉴리렌 이미징을 위한 랜덤 도트 배열 투영용 이진 회절광학소자를 설계하고 제작하였다. 이 연구에서 적용된 회절광학소자는 단 두 단계의 위상 및 10 ㎛의 피치를 갖는 이진 위상 회절 격자로, 제작 단가 및 제작 공정의 용이성을 위하여 선택되었다. 회절광학소자의 설계는 최종 패턴의 밝기 정보를 목적 함수로 사용하는 iterative Fourier transform algorithm을 적용하였다. 먼저 균일 밝기의 랜덤 도트 이미지를 생성하였고, 이를 최종 목표 이미지(패턴)로 적용한 결과, 위치(시야)에 따른 랜덤 도트의 밝기 변화를 확인하였다. 이를 해결하기 위하여 최종 목표 패턴에 가우시안 가중치를 적용한 개선 설계를 적용하였고, 그 결과 패턴 밝기 균일도를 52.7%에서 90.8%까지 향상시켰다. 이후, 바이너리 회절 소자 및 이를 적용한 빔 투사기를 제작하여 설계 결과를 검증하였다. 검증 결과 투사 거리 5 m에서 설계 목표인 430 mm × 430 mm 투광면적, 10,000개 이상의 랜덤 도트 패턴의 생성을 확인하였다. 측정된 균일도는 시뮬레이션에서 예상되었던 균일도보다 다소 적은 84.5%이나, 이는 회절 격자 형상, 특히 모서리 뭉개짐 및 간격 오류에 의한 것으로 추정된다.

• Smart Structures and Systems
• /
• 제25권3호
• /
• pp.369-384
• /
• 2020

A wavefield sparse reconstruction technique based on compressed sensing is developed in this work to dramatically reduce the number of measurements. Firstly, a severely underdetermined representation of guided wavefield at a snapshot is established in the spatial domain. Secondly, an optimal compressed sensing model of guided wavefield sparse reconstruction is established based on l₁-norm penalty, where a suite of discrete cosine functions is selected as the dictionary to promote the sparsity. The regular, random and jittered undersampling schemes are compared and selected as the undersampling matrix of compressed sensing. Thirdly, a gradient projection method is employed to solve the compressed sensing model of wavefield sparse reconstruction from highly incomplete measurements. Finally, experiments with different excitation frequencies are conducted on an aluminum plate to verify the effectiveness of the proposed sparse reconstruction method, where a scanning laser Doppler vibrometer as the true benchmark is used to measure the original wavefield in a given inspection region. Experiments demonstrate that the missing wavefield data can be accurately reconstructed from less than 12% of the original measurements; The reconstruction accuracy of the jittered undersampling scheme is slightly higher than that of the random undersampling scheme in high probability, but the regular undersampling scheme fails to reconstruct the wavefield image; A quantified mapping relationship between the sparsity ratio and the recovery error over a special interval is established with respect to statistical modeling and analysis.

• 제어로봇시스템학회:학술대회논문집
• /
• 제어로봇시스템학회 2004년도 ICCAS
• /
• pp.1491-1495
• /
• 2004

For the evaluation of the mobile robot performance in complex environments, the experimental approach in an actual physical environment has been commonly taken. In the physical experimental approach, however, it is quite difficult to define the proper environment for the evaluation due to the lack of commonly agreed characteristics of the test environment. Particularly the number of combinations of types and physical parameters of the obstacles that the mobile robot is expected to deal with is practically unlimited. In an effort to simplify and improve the effectiveness of the evaluation process, we propose an evaluation method using decomposed environmental elements, where we evaluated the performance of the robot for a small group of simple and decomposed obstacle components, for examples projection and slope, instead of a large group of complicated random obstacles. The paper describes a set of simple obstacle models and performance parameters that we have chosen for the effective evaluation process. As an alternative to the physical experimental evaluation approach, in this paper, we used a virtual evaluation environment where the robot and the physical test environment has been modeled using a commercial multi-body dynamics analysis packaged called RecurDyn.

• 한국소음진동공학회:학술대회논문집
• /
• 한국소음진동공학회 2002년도 추계학술대회논문집
• /
• pp.410-415
• /
• 2002

In this article, spatial filtering procedures with application to spherical acoustical holography are discussed. Planar and cylindrical holography are the most widely used amongst the various nearfield acoustical holography techniques. However, when the geometry of a source is similar to a sphere, spherical holography may yield better results than other types of holography since there are no errors due to truncation of the sound field in the spherical case. Spatial filtering affects the accuracy of spherical acoustical holography critically, especially in the case of backward projection. Thus spatial filtering is essential for successful application of spherical holography. In the present work, various filtering methods were evaluated in simulations made using sound pressure fields of various types and with different levels of random spatial noise. It was found that a procedure based on eliminating spherical harmonic coefficients that contribute insignificantly to the total sound power of the source gave the best results on average of the different procedures considered here. Spherical holography procedures were also verified experimentally. Reliable results were obtained using the power filtering algorithm. Thus it was concluded that spherical holography combined with power filtering may prove to be a useful tool for noise source identification.

• 대한치과기공학회지
• /
• 제36권3호
• /
• pp.205-217
• /
• 2014

Purpose: This study will examine the differences among the college dental morphology textbooks in light of their contents and learning objectives through which we will propose an optimal way of consolidating those differences found. Methods: Five college textbooks adopted in the dental related departments were selected by random and the overview and subdivisions of contents were compared and closely analysed with regards to the learning objectives. Results: Firstly, all of the dental morphology textbooks cover the learning objectives of the dental morphology subject, especially in the area of the overview of dental morphology, the permanent tooth, deciduous tooth. Only the dentistry textbooks explain the learning objective of the occlusion. Secondly, differences in content were found in the area of component tissue and around tissue, dental formula of deciduous teeth, spinous process, buccal pit, enamel projection, curve symbol, tip of cusp position of proximal surface of permanent mandibular canines, buccal cusp position of permanent mandibular second premolars. Conclusion: It is imperative to delineate some meaningful and critical differences in contents among the dental morphology textbooks and reflect this to each and every textbook to be published as a supplementary information guide or index.

• 한국정보통신학회:학술대회논문집
• /
• 한국해양정보통신학회 2003년도 추계종합학술대회
• /
• pp.797-800
• /
• 2003

본 논문은 차량 번호판 중앙부 위치값을 기반으로한 신경망을 이용하여 차량의 번호판 영역을 추출하는 방법을 제안하고자 한다. 임의의 숫자들로 정의된 표시영역에 대한 학습패턴과 넓은 범위를 수용할 수 있도록 한 신경망의 학습패턴을 이용하여 보다 효율적인 방법을 제시하였다. 학습패턴으로 차량 번호판 인식의 최적화을 이루었고 차량번호 및 헤드라이트 부분의 은닉효과와, 학습패턴의 확대 및 감소에 대하여 연구하였단. 위의 과정을 통하여 지하주차장에서 595여대의 자동차에 대하여 번호판 영역을 추출한 결과 98.5%의 인식율을 보여주었다.