http://chineseinput.net/에서 pinyin(병음)방식으로 중국어를 변환할 수 있습니다.
변환된 중국어를 복사하여 사용하시면 됩니다.
선호도 기반 최단경로 탐색을 위한 휴리스틱 융합 알고리즘
옥승호(Seung-Ho Ok),안진호(Jin-Ho Ahn),강성호(Sungho Kang),문병인(Byungin Moon) 大韓電子工學會 2010 電子工學會論文誌-TC (Telecommunications) Vol.47 No.8
본 논문에서는 개미 군집 최적화 (Ant Colony Optimization; ACO) 및 A* 휴리스틱 알고리즘이 융합된 선호도 기반 경로탐색 알고리즘을 제안한다. 최근 ITS (Intelligent Transportation Systems)의 개발과 함께 차량용 내비게이션의 사용이 증가하면서 경로탐색 알고리즘의 중요성이 더욱 높아지고 있다. 기존의 Dijkstra 및 A*와 같은 대부분의 최단경로 탐색 알고리즘은 최단거리 또는 최단시간 경로 탐색을 목표로 한다. 하지만 이러한 경로 탐색 결과는 더 안전하고 특정 경로를 선호하는 운전자를 위한 최적의 경로가 아니다. 따라서 본 논문에서는 선호도 기반 최단 경로 탐색 알고리즘을 제안한다. 제안된 알고리즘은 주어진 맵의 링크 속성 정보를 이용하며, 각 링크에 대한 사용자 선호도는 내비게이션 사용자에 의해 설정되어 진다. 제안된 알고리즘은 C로 구현하였으며, 64노드 및 118링크로 구성된 맵에서 다양한 파라미터를 통해 성능을 측정한 결과 본 논문에서 제안한 휴리스틱 융합 알고리즘은 선호도 기반 경로뿐만 아니라 최단 경로 탐색에도 적합함을 알 수 있었다. In this paper, we propose a preference-based shortest path algorithm which is combined with Ant Colony Optimization (ACO) and A* heuristic algorithm. In recent years, with the development of ITS (Intelligent Transportation Systems), there has been a resurgence of interest in a shortest path search algorithm for use in car navigation systems. Most of the shortest path search algorithms such as Dijkstra and A* aim at finding the distance or time shortest paths. However, the shortest path is not always an optimum path for the drivers who prefer choosing a less short, but more reliable or flexible path. For this reason, we propose a preference-based shortest path search algorithm which uses the properties of the links of the map. The preferences of the links are specified by the user of the car navigation system. The proposed algorithm was implemented in C and experiments were performed upon the map that includes 64 nodes with 118 links. The experimental results show that the proposed algorithm is suitable to find preference-based shortest paths as well as distance shortest paths.
고속 스테레오 정합을 위한 점진적 시차 탐색 범위 추정 기법
옥승호(Seung-Ho Ok),문병인(Byungin Moon) 한국정보기술학회 2017 한국정보기술학회논문지 Vol.15 No.1
In this paper, to reduce the computational overhead of the conventional area-based stereo matching algorithm, we propose a progressive disparity search range estimation method for a high-speed stereo matching algorithm. The proposed method can increase the matching accuracy by avoiding the matching on unnecessary areas and decrease the computational overhead by using the proposed disparity search range estimation method that uses the previously obtained depth map. The experimental results showed that the average matching accuracy was increased by 10% and the average computation overhead was reduced by 62% compared with the previously proposed stereo matching algorithms. This is because the disparity search range was decreased by 81% so that the proposed algorithm can avoid unnecessary matching areas.
김태원,옥승호,허경용,이임건,Kim, Tae-Won,Ok, Seung-Ho,Heo, Gyeongyong,Lee, Imgeun 한국정보통신학회 2020 한국정보통신학회논문지 Vol.24 No.7
최근 2층 버스 등 전고가 높은 차량이 증가함에 따라 지정된 경로 이탈 및 운전자 부주의로 인해 교량 및 터널 등에서 차량 상부 충돌 사고가 발생하고 있다. 기존 전방 충돌 경고 시스템의 경우 차량 및 보행자 등에 한정되어 경고를 발생하기 때문에 전고가 높은 차량을 위한 통과 높이 경고 시스템으로는 사용이 어렵다. 이에 본 논문에서는 복수개의 라이다 센서를 사용하여 세그먼트별 데이터의 상관도 및 시계열 특성을 판단한 후 차량 상부 충돌 가능성을 미리 판단하여 경고를 발생시키는 시스템을 제안한다. 또한, 제안하는 시스템은 실도로 주행 테스트 및 한국 자동차 안전 연구원에서 시스템 성능 평가를 통해 정상 동작을 확인하였다. Recently, as the number of high-height vehicles such as double-decker buses has increased, collision accidents have occurred in bridges and tunnels due to the deviation from the designated routes and driver's carelessness. In the case of the existing front collision warning system, it is limited to vehicles and pedestrians, so it is difficult to use it as a pass height warning system for the high height vehicles. In this paper, we propose a system that generates a warning by determining the correlation and time series characteristics of data for each segment using multiple lidar sensors and then determining the possibility of collision in the upper part of the vehicle. Also, the proposed system confirmed the proper operation through a real-time driving test and a system performance evaluation by the Korea Automobile Testing & Research Institute.
합성곱 신경망의 추론 정확도 향상을 위한 하드웨어 구조 연구
전성호(Seong-Ho Jeon),옥승호(Seung-Ho Ok) 한국정보기술학회 2021 한국정보기술학회논문지 Vol.19 No.12
In this paper, we propose a hardware architecture with built-in batch normalization to improve inference accuracy of convolutional neural networks. Batch normalization improves inference accuracy by normalizing the data distribution by the convolutional neural network layer. However, operations such as average and variance are required, and if there is a restriction on hardware resource usage, the operation accuracy decreases. Therefore, this paper proposes a hardware architecture in which batch normalization operations can be omitted by applying an 8-bit quantization technique when inferencing to improve computational accuracy and utilize internal control signals. The proposed hardware architecture was modeled using Verilog-HDL. Then the intermediate and final output values of the hardware operation were compared and evaluated with the output results of the software-based verification model and synthesized using the Samsung 28-nm process.
잡음에 강인한 스테레오 카메라 기반 ROI 알고리즘에 관한 연구
이상호(Sang-Ho Lee),옥승호(Seung-Ho Ok) 한국정보기술학회 2020 한국정보기술학회논문지 Vol.18 No.12
In this paper, we propose stereo camera-based noise-robust Region Of Interest(ROI) algorithm. In the case of the existing stereo camera-based ROI algorithm, some noise due to illumination and disturbance was removed by using binarization and morphology calculations in the process of acquiring the depth image, but shows a low ROI detection rate. In order to solve this problem, the proposed ROI algorithm proposes a mode detection algorithm through correlation analysis of labeling coordinates for sequential frames of preprocessed depth images, and a quick-sort technique is applied to reduce the computational amount of the mode detection algorithm. By applying the mode detection algorithm in the existing ROI detection algorithm, the proposed ROI algorithm showed a higher ROI detection rate in a complex background than the existing ROI algorithm. As a result of performance comparison and evaluation, the average detection rate of the proposed ROI algorithm improved by about 29% compared to the existing method.