• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.25 no.7
• /
• pp.962-970
• /
• 2021

The cause of the majority of vehicle accidents is a safety issue due to the driver's inattention, such as drowsy driving. A forward collision warning system (FCWS) can significantly reduce the number and severity of accidents by detecting the risk of collision with vehicles in front and providing an advanced warning signal to the driver. This paper describes a low power embedded system based FCWS for safety. The algorithm computes time to collision (TTC) through detection, tracking, distance calculation for the vehicle ahead and current vehicle speed information with a single camera. Additionally, in order to operate in real time even in a low-performance embedded system, an optimization technique in the program with high and low levels will be introduced. The system has been tested through the driving video of the vehicle in the embedded system. As a result of using the optimization technique, the execution time was about 170 times faster than that when using the previous non-optimized process.

• Proceedings of the Korean Society of Machine Tool Engineers Conference
• /
• 2002.04a
• /
• pp.393-398
• /
• 2002

Neural networks are used in the framework of sensorbased tracking control of robot manipulators. They learn by practice movements the relationship between PSD (an analog Position Sensitive Detector) sensor readings for target positions and the joint commands to reach them. Using this configuration, the system can track or follow a moving or stationary object in real time. forthermore, an efficient neural network architecture has been developed for real time learning. This network uses multiple sets of simple backpropagation networks one of which is selected according to which division (corresponding to a cluster of the self-organizing feature map) in data space the current input data belongs to. This lends itself to a very training and processing implementation required for real time control.

• Proceedings of the Korean Society for Agricultural Machinery Conference
• /
• 1999.07a
• /
• pp.142-149
• /
• 1999

최근 주변첨단기술이 발달하여 수확, 접목 및 착유 등 농작업에 로봇을 응용하기 위한 연구가 수행되고 있다. 작업의 특성에 따라 응용되는 로봇기술도 차이는 있겠으나 수확용 로봇의 경우 대상물을 인식하고 위치를 검출하는 기술개발이 선행되어야만 대상물을 파지 또는 절단하여 수확이 가능하며, 대상물의 인식 및 위치 검출에는 영상처리기술이 많이 응용되고 있다. 따라서 수확로봇의 작업성능은 대상물을 얼마나 정확하게 인식하고 위치를 정확하게 검출하는 것이 중요하다. 그러나 작업대상이 일정한 조건에서 표준화 되어있는 일반 산업분야와는 달리 농산물의 수확작업현장은 대상물의 배치, 크기 및 형상 둥이 일정하지 않고 작업환경 또한 다양하여 외란의 영향을 많기 받기 때문에 시각장치의 개발은 매우 중요하다. (중략)

• Proceedings of the Korea Information Processing Society Conference
• /
• 2003.11a
• /
• pp.53-56
• /
• 2003

본 논문에서는 비디오 지리정보시스템 구축을 위해 특징점을 이용한 객체 추적 알고리즘을 제안한다. 기존의 특징점을 이용한 객체 추적 알고리즘은 카메라의 움직임이 배제된, 즉 배경이 고정된 영상에서 객체의 특성을 이용한 방식으로, 운행중인 차량에 부착된 카메라에서 획득한 영상에서는 배경이 움직이므로 객체를 정확하게 추적하지 못하는 문제점이 있다. 본 논문에서는 이러한 문제점을 개선하기 위해 연속된 카메라 영상에서 모션벡터를 이용해 시간에 따른 객체의 변화율을 유도하고 유도된 변화율을 이용하여 객체 추적 알고리즘을 제안한다.

• Proceedings of the IEEK Conference
• /
• 2008.06a
• /
• pp.657-658
• /
• 2008

개인과 사회의 안전을 중요시하는 요즘 영상감시시스템의 비중은 날이 갈수록 커져가고 있다. 본 논문은 지능형 영상 감시 시스템에 적용되는 비전기반의 물체 추적시스템에 관한 내용으로 이루어져 있다. 지능형 영상 감시 시스템은 실제 국내외 많은 기업에서 활발히 연구하고 대표적인 분야이다. 하지만, 제품화되기에는 아직 많은 문제가 존재한다. 이러한 문제가 나타나는 가장 큰 원인 중 하나는 타겟의 크기가 변화하여 정보가 손상되는 경우이다. 이로 인해 정확한 타겟정보를 얻지 못하고, 배경정보를 갱신함으로써 추적성능을 크게 저하시키게 된다. 따라서 본 논문에서는 이러한 문제를 최소화하기 위하 방법을 제안한다.

• Proceedings of the Korean Society of Machine Tool Engineers Conference
• /
• 2001.04a
• /
• pp.239-243
• /
• 2001

Neural networks are used in the framework of sensorbased tracking control of robot manipulators. They learn by practice movements the relationship between PSD (an analog Position Sensitive Detector) sensor readings for target positions and the joint commands to reach them. Using this configuration, the system can track or follow a moving or stationary object in real time. Furthermore, an efficient neural network architecture has been developed for real time learning. This network uses multiple sets of simple backpropagation networks one of which is selected according to which division (corresponding to a cluster of the self-organizing feature map) in data space the current input data belongs to. This lends itself to a very training and processing implementation required for real time control.

• Proceedings of the Korean Society of Machine Tool Engineers Conference
• /
• 2000.04a
• /
• pp.171-175
• /
• 2000

Neural networks are in the framework of sensorbased tracking control of robot manipulators. They learn by practice movements the relationship between PSD (an analog Position Sensitive Detector) sensor readings for target positions and the joint commands to reach them. Using this configuration, the system can track or follow a moving or stationary object in real time. Furthermore, an efficient neural network architecture has been developed for real time learning. This network uses multiple sets of simple backpropagation networks one of which is selected according to which division (corresponding to a cluster of the self-organizing feature map) in data space the current input data belongs to. This lends itself to a very training and processing implementation required for real time control.

• The Journal of Korean Institute of Communications and Information Sciences
• /
• v.15 no.8
• /
• pp.708-714
• /
• 1990

This paper has proposed a self-tuning controller for tracking reference trajectory by measuring End-point of arm on robot manipulator whose link is light and flexibls, and proved the perforformance of the algorithm proposed through the computer simulation. As an object of control, a flexible robot manipulator with two-links was selected. As for structure of model, it utilized an assume mode shape method with include travity force and derived a dynaics equation by adapting two kinds of vibration mode of each.

• Proceedings of the Korean Institute of Intelligent Systems Conference
• /
• 2008.04a
• /
• pp.143-146
• /
• 2008

본 논문에서는 군집로봇시스템에서 목표물 추적을 위하여 SVM을 이용한 12각형 기반의 Q-learning 알고리즘을 제안한다. 제안한 알고리즘의 유효성을 보이기 위해 본 논문에서는 여러대의 로봇과 장애물 그리고 하나의 목표물을 정하고, 각각의 로봇이 숨겨진 목표물을 찾아내는 실험을 가정하여 무작위, DBAM과 ABAM의 융합 모델, 그리고 마지막으로 본 논문에서 제안한 SVM과 12각형 기반의 Q-learning 알고리즘을 이용하여 실험을 수행하고, 이 3가지 방법을 비교하여 본 논문의 유효성을 검증하였다.

• Proceedings of the Korean Society of Machine Tool Engineers Conference
• /
• 2000.10a
• /
• pp.108-112
• /
• 2000

Neural networks are used in the framework of sensorbased tracking control of robot manipulators. They learn by practice movements the relationship between PSD (an analog Position Sensitive Detector) sensor readings for target positions and the joint commands to reach them. Using this configuration, the system can track or follow a moving or stationary object in real time. Furthermore, an efficient neural network architecture has been developed for real time learning. This network uses multiple sets of simple backpropagation networks one of which is selected according to which division (corresponding to a cluster of the self-organizing feature map) in data space the current input data belongs to. This lends itself to a very training and processing implementation required for real time control.