• Proceedings of the Korean Information Science Society Conference
• /
• 2012.06b
• /
• pp.360-362
• /
• 2012

교통수단의 발달과 생활수준의 향상으로 도로에 차량이 많이 늘어나고 교통사고가 많이 발생함에 따라, 교통사고 자동인식 시스템에 관한 연구가 많이 진행되고 있다. 본 논문에서는 카메라의 위치에 따라 두 객체의 관심영역 사이의 겹침을 해석하는 것이 달라져 규칙이 변하는 것을 방지하고, 사람의 추론과정과 같이 교통사고를 퍼지 규칙으로 모델링하여 획득한 데이터가 부정확할 경우에 발생하는 잘못된 추론을 보정하기 위한 퍼지 규칙기반 시스템을 제안한다. 카이스트 삼거리에서 촬영한 9개의 사고 시나리오 데이터에 대해 실험하여 DR 87.34%, CDR 89.13%, FAR 10.75%의 결과를 얻었고, 이를 기존의 규칙기반 시스템, 규칙-확률 시스템과 비교하였다.

• Proceedings of the Korean Information Science Society Conference
• /
• 2012.06b
• /
• pp.351-353
• /
• 2012

교통 표지판은 도로 주행에 있어 분별력 있는 정보를 제공한다. 하지만 주행 중에 가로수나 다른 자동차에 의해 교통 표지판은 가려져 있거나 훼손된 경우가 많다. 또한 자동차가 커브할 때 카메라 영상에는 회전된 객체로 보이게 된다. 이런 경우에 교통 표지판의 인식이 어렵기 때문에 본 논문에서는 각 문제점에 모두 강건한 피처를 이용해 매칭하는 방법을 제안하였다. 본 논문에서 제안한 방법에 기반하여 주행 중 영상에서 보다 분별력 있는 정보를 획득하여 더 많은 응용 분야에 적용할 것으로 기대한다.

• Proceedings of the Korean Information Science Society Conference
• /
• 2000.10a
• /
• pp.89-91
• /
• 2000

최근 들어 멀티미디어와 같은 고차원 데이터를 효율적으로 처리하기 위한 고차원 인덱싱구조에 대한 연구가 활발히 진행되어 왔으며, 특히 R-tree를 기반으로 하는 인덱싱 구조가 가장 많이 발표되었다. 그러나, R-tree 계열의 색인기법은 데이터 삽입시 삽입순서를 비효율적으로 배정하는 경우 실제적으로 자신과 거리가 먼 객체들과 동일한 노드에 삽입될 수 있다. 이는 인덱싱 구조내에 많은 겹침을 초래하고 결과적으로 검색성능을 저하시킬 수 있다. 본 논문은 이러한 단점을 극복하기 위하여 데이터가 지니는 특성 중, 공간적인 위치정보를 분석하여 클러스터를 형성하고 이를 이용하여 인덱싱 형성에 적용하였고, 질의 처리시 검색성능이 향상되도록 하였다. 또한 실험에서는 실제 이미지 데이터에 대하여 제안한 기법을 적용하여 성능이 향상되었음을 나타내었다.

• Proceedings of the Korean Information Science Society Conference
• /
• 2002.10c
• /
• pp.28-30
• /
• 2002

시간에 따라 연속적으로 위치가 변화하는 객체를 이동체라 한다. 기존의 R-Tree를 사용한 이동체 색인에 관한 연구에서는 현재 위치 질의 시 고비용의 연산이 요구되고, 시간축의 값이 증가하는 방향으로 보고되는 이동체의 위치데이터의 특징을 고려한 노드 분할 정책이 제안되지 않았다. 이 논문에서는 이동체의 현재 위치 및 과거 위치에 대한 색인 방법인 CPTR-Tree(Current Position and Trajectory R-Tree)를 제안한다. 특히, 제안 방법에서 이동체의 현재 위치에 대한 공간차원의 PMBR(Point MBR)을 유지함으로써, 현재 위치 질의 처리시 불필요한 노드 접근 횟수를 줄일 수 있어 성능향상을 할 수 있다. 그리고, 시간축의 값이 증가하는 형태로 보고되는 이동체 위치 데이터의 특징을 고려하여 시간축 분할시 SP(Split Parameter) 분할 방법을 제공함으로써 노드 공간 활용률을 높여 색인의 크기를 줄이고, 공간축 분할시 노드 겹침을 줄이는 동적 클리핑 분할 정책을 제시하여 이동체 과거 위치 검색 효율을 높인다.

• Journal of IKEEE
• /
• v.28 no.1
• /
• pp.65-71
• /
• 2024

This paper proposes a neural network that can reduce the number of parameters while reducing matching errors in occluded regions to increase the accuracy of depth maps in stereo matching. Stereo matching-based object recognition is utilized in many fields to more accurately recognize situations using images. When there are many objects in a complex image, an occluded area is generated due to overlap between objects and occlusion by background, thereby lowering the accuracy of the depth map. To solve this problem, existing research methods that create context information and combine it with the cost volume or RoIselect in the occluded area increase the complexity of neural networks, making it difficult to learn and expensive to implement. In this paper, we create a depthwise seperable neural network that enhances regional feature extraction before cost volume generation, reducing the number of parameters and proposing a neural network that is robust to occlusion errors. Compared to PSMNet, the proposed neural network reduced the number of parameters by 30%, improving 5.3% in color error and 3.6% in test loss.

• Journal of KIISE:Software and Applications
• /
• v.32 no.5
• /
• pp.449-460
• /
• 2005

This paper proposes a method to track persons robustly in illumination changes and partial occlusions in color video frames acquired from a fixed camera. To solve a problem of changing appearance by illumination change, a time-independent intrinsic image is used to remove noises in an frame and is adaptively updated frame-by-frame. We use a hierarchical human model including body color information in order to track persons in occlusion. The tracked human model is recorded into a persons' list for some duration after the corresponding person's exit and is recovered from the list after her reentering. The proposed method was experimented in several indoor and outdoor scenario. This demonstrated the potential effectiveness of an adaptive model-base method that corrected distorted person's color information by lighting changes, and succeeded tracking of persons which was overlapped in a frame.

• Journal of KIISE:Databases
• /
• v.35 no.5
• /
• pp.432-446
• /
• 2008

Queries for tracing tag locations are among the most challenging requirements in RFID based applications, including automated manufacturing, inventory tracking and supply chain management. For efficient query processing, a previous study proposed the index scheme for storing tag objects, based on the moving object index, in 3-dimensional domain with the axes being the tag identifier, the reader identifier, and the time. In a different way of a moving object index, the ranges of coordinates for each domain are quite different so that the distribution of query regions is skewed to the reader identifier domain. Previous indexes for tags, however, do not consider the skewed distribution for query regions. This results in producing many overlaps between index nodes and query regions and then causes the problem of traversing many index nodes. To solve this problem, we propose a new disproportional insertion and split policy of the index for RFID tags which is based on the R*-tree. For efficient insertion of tag data, our method derives the weighted margin for each node by using weights of each axis and margin of nodes. Based the weighted margin, we can choose the subtree and the split method in order to insert tag data with the minimum cost. Proposed insertion method also reduces the cost of region query by reducing overlapped area of query region and MBRs. Our experiments show that the index based on the proposed insertion and split method considerably improves the performance of queries than the index based on the previous methods.

• Proceedings of the Korean Institute of Navigation and Port Research Conference
• /
• 2010.04a
• /
• pp.139-142
• /
• 2010

ENC(Electronic Navigational Chart) is an electronic chart that digitize nautical chart including all kinds of information of nautical chart. Mariners have difficulty reading information because of clutter problem if display scale of ECDIS is lower than compilation scale of ENCs. IHO made SCAMIN improvement model included in S-65 Standard to settle this ENC clutter problem. This SCAMIN model includes both simple restriction that should safety some requirement and share geographic information. and hard restriction that should overlap with area object such as depth area, land area, dredge area. It also apply SCAMIN step of sounding value according to importance of safety navigation. In this paper, we analyzed the restrictions for SCAMIN model of S-65 Standard, developed algorithm and application to apply theses restrictions reasonably and mechanically. Also we applied this algorithm and application to Korea ENCs and evaluated the result of performance.

• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.14 no.11
• /
• pp.2403-2412
• /
• 2010

ENC(Electronic Navigational Chart) is an electronic chart including all kinds of information of nautical chart. Mariners have difficulty reading information because of clutter problem if display scale of ECDIS is lower than compilation scale of ENCs. IRO made SCAMIN improvement model included in S-65 Standard to settle this ENC clutter problem. This SCAMIN model includes both simple restriction that should satisfy some requirement and share geographic information, and hard restriction that should overlap with area object such as depth area, land area, dredge area. It also apply SCAMIN step of sounding value according to importance of safety navigation. In this paper, we analyzed the restrictions for SCAMIN model of S-65 Standard, developed algorithm and application to apply theses restrictions reasonably and mechanically. Also we applied this algorithm and application to Korea ENCs and evaluated the result of performance.

• Journal of the Institute of Electronics and Information Engineers
• /
• v.51 no.2
• /
• pp.69-75
• /
• 2014

Existing People Counting System using a single overhead mounted camera has limitation in object recognition and counting in various environments. Those limitations are attributable to overlapping, occlusion and external factors, such as over-sized belongings and dramatic light change. Thus, this paper proposes the new concept of People Counting System by Facial Age Group using two depth cameras, at overhead and frontal viewpoints, in order to improve object recognition accuracy and robust people counting to external factors. The proposed system is counting the pedestrians by five process such as overhead image processing, frontal image processing, identical object recognition, facial age group classification and in-coming/out-going counting. The proposed system developed by C++, OpenCV and Kinect SDK, and it target group of 40 people(10 people by each age group) was setup for People Counting and Facial Age Group classification performance evaluation. The experimental results indicated approximately 98% accuracy in People Counting and 74.23% accuracy in the Facial Age Group classification.